OpenCloudOS-Kernel/mm/readahead.c

607 lines
16 KiB
C
Raw Normal View History

/*
* mm/readahead.c - address_space-level file readahead.
*
* Copyright (C) 2002, Linus Torvalds
*
* 09Apr2002 Andrew Morton
* Initial version.
*/
#include <linux/kernel.h>
#include <linux/dax.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/gfp.h>
#include <linux/export.h>
#include <linux/blkdev.h>
#include <linux/backing-dev.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/pagevec.h>
#include <linux/pagemap.h>
#include <linux/syscalls.h>
#include <linux/file.h>
#include <linux/mm_inline.h>
#include <linux/blk-cgroup.h>
#include <linux/fadvise.h>
#include "internal.h"
/*
* Initialise a struct file's readahead state. Assumes that the caller has
* memset *ra to zero.
*/
void
file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping)
{
ra->ra_pages = inode_to_bdi(mapping->host)->ra_pages;
ra->prev_pos = -1;
}
EXPORT_SYMBOL_GPL(file_ra_state_init);
/*
* see if a page needs releasing upon read_cache_pages() failure
* - the caller of read_cache_pages() may have set PG_private or PG_fscache
* before calling, such as the NFS fs marking pages that are cached locally
* on disk, thus we need to give the fs a chance to clean up in the event of
* an error
*/
static void read_cache_pages_invalidate_page(struct address_space *mapping,
struct page *page)
{
if (page_has_private(page)) {
if (!trylock_page(page))
BUG();
page->mapping = mapping;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
do_invalidatepage(page, 0, PAGE_SIZE);
page->mapping = NULL;
unlock_page(page);
}
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
put_page(page);
}
/*
* release a list of pages, invalidating them first if need be
*/
static void read_cache_pages_invalidate_pages(struct address_space *mapping,
struct list_head *pages)
{
struct page *victim;
while (!list_empty(pages)) {
victim = lru_to_page(pages);
list_del(&victim->lru);
read_cache_pages_invalidate_page(mapping, victim);
}
}
/**
* read_cache_pages - populate an address space with some pages & start reads against them
* @mapping: the address_space
* @pages: The address of a list_head which contains the target pages. These
* pages have their ->index populated and are otherwise uninitialised.
* @filler: callback routine for filling a single page.
* @data: private data for the callback routine.
*
* Hides the details of the LRU cache etc from the filesystems.
*
* Returns: %0 on success, error return by @filler otherwise
*/
int read_cache_pages(struct address_space *mapping, struct list_head *pages,
int (*filler)(void *, struct page *), void *data)
{
struct page *page;
int ret = 0;
while (!list_empty(pages)) {
page = lru_to_page(pages);
list_del(&page->lru);
mm, fs: obey gfp_mapping for add_to_page_cache() Commit 6afdb859b710 ("mm: do not ignore mapping_gfp_mask in page cache allocation paths") has caught some users of hardcoded GFP_KERNEL used in the page cache allocation paths. This, however, wasn't complete and there were others which went unnoticed. Dave Chinner has reported the following deadlock for xfs on loop device: : With the recent merge of the loop device changes, I'm now seeing : XFS deadlock on my single CPU, 1GB RAM VM running xfs/073. : : The deadlocked is as follows: : : kloopd1: loop_queue_read_work : xfs_file_iter_read : lock XFS inode XFS_IOLOCK_SHARED (on image file) : page cache read (GFP_KERNEL) : radix tree alloc : memory reclaim : reclaim XFS inodes : log force to unpin inodes : <wait for log IO completion> : : xfs-cil/loop1: <does log force IO work> : xlog_cil_push : xlog_write : <loop issuing log writes> : xlog_state_get_iclog_space() : <blocks due to all log buffers under write io> : <waits for IO completion> : : kloopd1: loop_queue_write_work : xfs_file_write_iter : lock XFS inode XFS_IOLOCK_EXCL (on image file) : <wait for inode to be unlocked> : : i.e. the kloopd, with it's split read and write work queues, has : introduced a dependency through memory reclaim. i.e. that writes : need to be able to progress for reads make progress. : : The problem, fundamentally, is that mpage_readpages() does a : GFP_KERNEL allocation, rather than paying attention to the inode's : mapping gfp mask, which is set to GFP_NOFS. : : The didn't used to happen, because the loop device used to issue : reads through the splice path and that does: : : error = add_to_page_cache_lru(page, mapping, index, : GFP_KERNEL & mapping_gfp_mask(mapping)); This has changed by commit aa4d86163e4 ("block: loop: switch to VFS ITER_BVEC"). This patch changes mpage_readpage{s} to follow gfp mask set for the mapping. There are, however, other places which are doing basically the same. lustre:ll_dir_filler is doing GFP_KERNEL from the function which apparently uses GFP_NOFS for other allocations so let's make this consistent. cifs:readpages_get_pages is called from cifs_readpages and __cifs_readpages_from_fscache called from the same path obeys mapping gfp. ramfs_nommu_expand_for_mapping is hardcoding GFP_KERNEL as well regardless it uses mapping_gfp_mask for the page allocation. ext4_mpage_readpages is the called from the page cache allocation path same as read_pages and read_cache_pages As I've noticed in my previous post I cannot say I would be happy about sprinkling mapping_gfp_mask all over the place and it sounds like we should drop gfp_mask argument altogether and use it internally in __add_to_page_cache_locked that would require all the filesystems to use mapping gfp consistently which I am not sure is the case here. From a quick glance it seems that some file system use it all the time while others are selective. Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: Dave Chinner <david@fromorbit.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Ming Lei <ming.lei@canonical.com> Cc: Andreas Dilger <andreas.dilger@intel.com> Cc: Oleg Drokin <oleg.drokin@intel.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-10-16 06:28:24 +08:00
if (add_to_page_cache_lru(page, mapping, page->index,
mm, memcg: use consistent gfp flags during readahead Vladimir has noticed that we might declare memcg oom even during readahead because read_pages only uses GFP_KERNEL (with mapping_gfp restriction) while __do_page_cache_readahead uses page_cache_alloc_readahead which adds __GFP_NORETRY to prevent from OOMs. This gfp mask discrepancy is really unfortunate and easily fixable. Drop page_cache_alloc_readahead() which only has one user and outsource the gfp_mask logic into readahead_gfp_mask and propagate this mask from __do_page_cache_readahead down to read_pages. This alone would have only very limited impact as most filesystems are implementing ->readpages and the common implementation mpage_readpages does GFP_KERNEL (with mapping_gfp restriction) again. We can tell it to use readahead_gfp_mask instead as this function is called only during readahead as well. The same applies to read_cache_pages. ext4 has its own ext4_mpage_readpages but the path which has pages != NULL can use the same gfp mask. Btrfs, cifs, f2fs and orangefs are doing a very similar pattern to mpage_readpages so the same can be applied to them as well. [akpm@linux-foundation.org: coding-style fixes] [mhocko@suse.com: restrict gfp mask in mpage_alloc] Link: http://lkml.kernel.org/r/20160610074223.GC32285@dhcp22.suse.cz Link: http://lkml.kernel.org/r/1465301556-26431-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov@parallels.com> Cc: Chris Mason <clm@fb.com> Cc: Steve French <sfrench@samba.org> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Jan Kara <jack@suse.cz> Cc: Mike Marshall <hubcap@omnibond.com> Cc: Jaegeuk Kim <jaegeuk@kernel.org> Cc: Changman Lee <cm224.lee@samsung.com> Cc: Chao Yu <yuchao0@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-27 06:24:53 +08:00
readahead_gfp_mask(mapping))) {
read_cache_pages_invalidate_page(mapping, page);
continue;
}
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
put_page(page);
ret = filler(data, page);
if (unlikely(ret)) {
read_cache_pages_invalidate_pages(mapping, pages);
break;
}
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
task_io_account_read(PAGE_SIZE);
}
return ret;
}
EXPORT_SYMBOL(read_cache_pages);
static int read_pages(struct address_space *mapping, struct file *filp,
mm, memcg: use consistent gfp flags during readahead Vladimir has noticed that we might declare memcg oom even during readahead because read_pages only uses GFP_KERNEL (with mapping_gfp restriction) while __do_page_cache_readahead uses page_cache_alloc_readahead which adds __GFP_NORETRY to prevent from OOMs. This gfp mask discrepancy is really unfortunate and easily fixable. Drop page_cache_alloc_readahead() which only has one user and outsource the gfp_mask logic into readahead_gfp_mask and propagate this mask from __do_page_cache_readahead down to read_pages. This alone would have only very limited impact as most filesystems are implementing ->readpages and the common implementation mpage_readpages does GFP_KERNEL (with mapping_gfp restriction) again. We can tell it to use readahead_gfp_mask instead as this function is called only during readahead as well. The same applies to read_cache_pages. ext4 has its own ext4_mpage_readpages but the path which has pages != NULL can use the same gfp mask. Btrfs, cifs, f2fs and orangefs are doing a very similar pattern to mpage_readpages so the same can be applied to them as well. [akpm@linux-foundation.org: coding-style fixes] [mhocko@suse.com: restrict gfp mask in mpage_alloc] Link: http://lkml.kernel.org/r/20160610074223.GC32285@dhcp22.suse.cz Link: http://lkml.kernel.org/r/1465301556-26431-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov@parallels.com> Cc: Chris Mason <clm@fb.com> Cc: Steve French <sfrench@samba.org> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Jan Kara <jack@suse.cz> Cc: Mike Marshall <hubcap@omnibond.com> Cc: Jaegeuk Kim <jaegeuk@kernel.org> Cc: Changman Lee <cm224.lee@samsung.com> Cc: Chao Yu <yuchao0@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-27 06:24:53 +08:00
struct list_head *pages, unsigned int nr_pages, gfp_t gfp)
{
struct blk_plug plug;
unsigned page_idx;
int ret;
blk_start_plug(&plug);
if (mapping->a_ops->readpages) {
ret = mapping->a_ops->readpages(filp, mapping, pages, nr_pages);
/* Clean up the remaining pages */
put_pages_list(pages);
goto out;
}
for (page_idx = 0; page_idx < nr_pages; page_idx++) {
struct page *page = lru_to_page(pages);
list_del(&page->lru);
mm, memcg: use consistent gfp flags during readahead Vladimir has noticed that we might declare memcg oom even during readahead because read_pages only uses GFP_KERNEL (with mapping_gfp restriction) while __do_page_cache_readahead uses page_cache_alloc_readahead which adds __GFP_NORETRY to prevent from OOMs. This gfp mask discrepancy is really unfortunate and easily fixable. Drop page_cache_alloc_readahead() which only has one user and outsource the gfp_mask logic into readahead_gfp_mask and propagate this mask from __do_page_cache_readahead down to read_pages. This alone would have only very limited impact as most filesystems are implementing ->readpages and the common implementation mpage_readpages does GFP_KERNEL (with mapping_gfp restriction) again. We can tell it to use readahead_gfp_mask instead as this function is called only during readahead as well. The same applies to read_cache_pages. ext4 has its own ext4_mpage_readpages but the path which has pages != NULL can use the same gfp mask. Btrfs, cifs, f2fs and orangefs are doing a very similar pattern to mpage_readpages so the same can be applied to them as well. [akpm@linux-foundation.org: coding-style fixes] [mhocko@suse.com: restrict gfp mask in mpage_alloc] Link: http://lkml.kernel.org/r/20160610074223.GC32285@dhcp22.suse.cz Link: http://lkml.kernel.org/r/1465301556-26431-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov@parallels.com> Cc: Chris Mason <clm@fb.com> Cc: Steve French <sfrench@samba.org> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Jan Kara <jack@suse.cz> Cc: Mike Marshall <hubcap@omnibond.com> Cc: Jaegeuk Kim <jaegeuk@kernel.org> Cc: Changman Lee <cm224.lee@samsung.com> Cc: Chao Yu <yuchao0@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-27 06:24:53 +08:00
if (!add_to_page_cache_lru(page, mapping, page->index, gfp))
mapping->a_ops->readpage(filp, page);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
put_page(page);
}
ret = 0;
out:
blk_finish_plug(&plug);
return ret;
}
/*
* __do_page_cache_readahead() actually reads a chunk of disk. It allocates
* the pages first, then submits them for I/O. This avoids the very bad
* behaviour which would occur if page allocations are causing VM writeback.
* We really don't want to intermingle reads and writes like that.
*
* Returns the number of pages requested, or the maximum amount of I/O allowed.
*/
unsigned int __do_page_cache_readahead(struct address_space *mapping,
struct file *filp, pgoff_t offset, unsigned long nr_to_read,
unsigned long lookahead_size)
{
struct inode *inode = mapping->host;
struct page *page;
unsigned long end_index; /* The last page we want to read */
LIST_HEAD(page_pool);
int page_idx;
unsigned int nr_pages = 0;
loff_t isize = i_size_read(inode);
mm, memcg: use consistent gfp flags during readahead Vladimir has noticed that we might declare memcg oom even during readahead because read_pages only uses GFP_KERNEL (with mapping_gfp restriction) while __do_page_cache_readahead uses page_cache_alloc_readahead which adds __GFP_NORETRY to prevent from OOMs. This gfp mask discrepancy is really unfortunate and easily fixable. Drop page_cache_alloc_readahead() which only has one user and outsource the gfp_mask logic into readahead_gfp_mask and propagate this mask from __do_page_cache_readahead down to read_pages. This alone would have only very limited impact as most filesystems are implementing ->readpages and the common implementation mpage_readpages does GFP_KERNEL (with mapping_gfp restriction) again. We can tell it to use readahead_gfp_mask instead as this function is called only during readahead as well. The same applies to read_cache_pages. ext4 has its own ext4_mpage_readpages but the path which has pages != NULL can use the same gfp mask. Btrfs, cifs, f2fs and orangefs are doing a very similar pattern to mpage_readpages so the same can be applied to them as well. [akpm@linux-foundation.org: coding-style fixes] [mhocko@suse.com: restrict gfp mask in mpage_alloc] Link: http://lkml.kernel.org/r/20160610074223.GC32285@dhcp22.suse.cz Link: http://lkml.kernel.org/r/1465301556-26431-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov@parallels.com> Cc: Chris Mason <clm@fb.com> Cc: Steve French <sfrench@samba.org> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Jan Kara <jack@suse.cz> Cc: Mike Marshall <hubcap@omnibond.com> Cc: Jaegeuk Kim <jaegeuk@kernel.org> Cc: Changman Lee <cm224.lee@samsung.com> Cc: Chao Yu <yuchao0@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-27 06:24:53 +08:00
gfp_t gfp_mask = readahead_gfp_mask(mapping);
if (isize == 0)
goto out;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
end_index = ((isize - 1) >> PAGE_SHIFT);
/*
* Preallocate as many pages as we will need.
*/
for (page_idx = 0; page_idx < nr_to_read; page_idx++) {
pgoff_t page_offset = offset + page_idx;
if (page_offset > end_index)
break;
page = xa_load(&mapping->i_pages, page_offset);
if (page && !xa_is_value(page)) {
/*
* Page already present? Kick off the current batch of
* contiguous pages before continuing with the next
* batch.
*/
if (nr_pages)
read_pages(mapping, filp, &page_pool, nr_pages,
gfp_mask);
nr_pages = 0;
continue;
}
mm, memcg: use consistent gfp flags during readahead Vladimir has noticed that we might declare memcg oom even during readahead because read_pages only uses GFP_KERNEL (with mapping_gfp restriction) while __do_page_cache_readahead uses page_cache_alloc_readahead which adds __GFP_NORETRY to prevent from OOMs. This gfp mask discrepancy is really unfortunate and easily fixable. Drop page_cache_alloc_readahead() which only has one user and outsource the gfp_mask logic into readahead_gfp_mask and propagate this mask from __do_page_cache_readahead down to read_pages. This alone would have only very limited impact as most filesystems are implementing ->readpages and the common implementation mpage_readpages does GFP_KERNEL (with mapping_gfp restriction) again. We can tell it to use readahead_gfp_mask instead as this function is called only during readahead as well. The same applies to read_cache_pages. ext4 has its own ext4_mpage_readpages but the path which has pages != NULL can use the same gfp mask. Btrfs, cifs, f2fs and orangefs are doing a very similar pattern to mpage_readpages so the same can be applied to them as well. [akpm@linux-foundation.org: coding-style fixes] [mhocko@suse.com: restrict gfp mask in mpage_alloc] Link: http://lkml.kernel.org/r/20160610074223.GC32285@dhcp22.suse.cz Link: http://lkml.kernel.org/r/1465301556-26431-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov@parallels.com> Cc: Chris Mason <clm@fb.com> Cc: Steve French <sfrench@samba.org> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Jan Kara <jack@suse.cz> Cc: Mike Marshall <hubcap@omnibond.com> Cc: Jaegeuk Kim <jaegeuk@kernel.org> Cc: Changman Lee <cm224.lee@samsung.com> Cc: Chao Yu <yuchao0@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-27 06:24:53 +08:00
page = __page_cache_alloc(gfp_mask);
if (!page)
break;
page->index = page_offset;
list_add(&page->lru, &page_pool);
if (page_idx == nr_to_read - lookahead_size)
SetPageReadahead(page);
nr_pages++;
}
/*
* Now start the IO. We ignore I/O errors - if the page is not
* uptodate then the caller will launch readpage again, and
* will then handle the error.
*/
if (nr_pages)
read_pages(mapping, filp, &page_pool, nr_pages, gfp_mask);
BUG_ON(!list_empty(&page_pool));
out:
return nr_pages;
}
/*
* Chunk the readahead into 2 megabyte units, so that we don't pin too much
* memory at once.
*/
int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
pgoff_t offset, unsigned long nr_to_read)
{
struct backing_dev_info *bdi = inode_to_bdi(mapping->host);
struct file_ra_state *ra = &filp->f_ra;
unsigned long max_pages;
if (unlikely(!mapping->a_ops->readpage && !mapping->a_ops->readpages))
return -EINVAL;
/*
* If the request exceeds the readahead window, allow the read to
* be up to the optimal hardware IO size
*/
max_pages = max_t(unsigned long, bdi->io_pages, ra->ra_pages);
nr_to_read = min(nr_to_read, max_pages);
while (nr_to_read) {
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
unsigned long this_chunk = (2 * 1024 * 1024) / PAGE_SIZE;
if (this_chunk > nr_to_read)
this_chunk = nr_to_read;
__do_page_cache_readahead(mapping, filp, offset, this_chunk, 0);
offset += this_chunk;
nr_to_read -= this_chunk;
}
return 0;
}
/*
* Set the initial window size, round to next power of 2 and square
* for small size, x 4 for medium, and x 2 for large
* for 128k (32 page) max ra
* 1-8 page = 32k initial, > 8 page = 128k initial
*/
static unsigned long get_init_ra_size(unsigned long size, unsigned long max)
{
unsigned long newsize = roundup_pow_of_two(size);
if (newsize <= max / 32)
newsize = newsize * 4;
else if (newsize <= max / 4)
newsize = newsize * 2;
else
newsize = max;
return newsize;
}
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
/*
* Get the previous window size, ramp it up, and
* return it as the new window size.
*/
static unsigned long get_next_ra_size(struct file_ra_state *ra,
unsigned long max)
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
{
unsigned long cur = ra->size;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
if (cur < max / 16)
return 4 * cur;
if (cur <= max / 2)
return 2 * cur;
return max;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
}
/*
* On-demand readahead design.
*
* The fields in struct file_ra_state represent the most-recently-executed
* readahead attempt:
*
* |<----- async_size ---------|
* |------------------- size -------------------->|
* |==================#===========================|
* ^start ^page marked with PG_readahead
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
*
* To overlap application thinking time and disk I/O time, we do
* `readahead pipelining': Do not wait until the application consumed all
* readahead pages and stalled on the missing page at readahead_index;
* Instead, submit an asynchronous readahead I/O as soon as there are
* only async_size pages left in the readahead window. Normally async_size
* will be equal to size, for maximum pipelining.
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
*
* In interleaved sequential reads, concurrent streams on the same fd can
* be invalidating each other's readahead state. So we flag the new readahead
* page at (start+size-async_size) with PG_readahead, and use it as readahead
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
* indicator. The flag won't be set on already cached pages, to avoid the
* readahead-for-nothing fuss, saving pointless page cache lookups.
*
* prev_pos tracks the last visited byte in the _previous_ read request.
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
* It should be maintained by the caller, and will be used for detecting
* small random reads. Note that the readahead algorithm checks loosely
* for sequential patterns. Hence interleaved reads might be served as
* sequential ones.
*
* There is a special-case: if the first page which the application tries to
* read happens to be the first page of the file, it is assumed that a linear
* read is about to happen and the window is immediately set to the initial size
* based on I/O request size and the max_readahead.
*
* The code ramps up the readahead size aggressively at first, but slow down as
* it approaches max_readhead.
*/
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
/*
* Count contiguously cached pages from @offset-1 to @offset-@max,
* this count is a conservative estimation of
* - length of the sequential read sequence, or
* - thrashing threshold in memory tight systems
*/
static pgoff_t count_history_pages(struct address_space *mapping,
pgoff_t offset, unsigned long max)
{
pgoff_t head;
rcu_read_lock();
head = page_cache_prev_miss(mapping, offset - 1, max);
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
rcu_read_unlock();
return offset - 1 - head;
}
/*
* page cache context based read-ahead
*/
static int try_context_readahead(struct address_space *mapping,
struct file_ra_state *ra,
pgoff_t offset,
unsigned long req_size,
unsigned long max)
{
pgoff_t size;
size = count_history_pages(mapping, offset, max);
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
/*
readahead: make context readahead more conservative This helps performance on moderately dense random reads on SSD. Transaction-Per-Second numbers provided by Taobao: QPS case ------------------------------------------------------- 7536 disable context readahead totally w/ patch: 7129 slower size rampup and start RA on the 3rd read 6717 slower size rampup w/o patch: 5581 unmodified context readahead Before, readahead will be started whenever reading page N+1 when it happen to read N recently. After patch, we'll only start readahead when *three* random reads happen to access pages N, N+1, N+2. The probability of this happening is extremely low for pure random reads, unless they are very dense, which actually deserves some readahead. Also start with a smaller readahead window. The impact to interleaved sequential reads should be small, because for a long run stream, the the small readahead window rampup phase is negletable. The context readahead actually benefits clustered random reads on HDD whose seek cost is pretty high. However as SSD is increasingly used for random read workloads it's better for the context readahead to concentrate on interleaved sequential reads. Another SSD rand read test from Miao # file size: 2GB # read IO amount: 625MB sysbench --test=fileio \ --max-requests=10000 \ --num-threads=1 \ --file-num=1 \ --file-block-size=64K \ --file-test-mode=rndrd \ --file-fsync-freq=0 \ --file-fsync-end=off run shows the performance of btrfs grows up from 69MB/s to 121MB/s, ext4 from 104MB/s to 121MB/s. Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Tested-by: Tao Ma <tm@tao.ma> Tested-by: Miao Xie <miaox@cn.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-12 05:21:47 +08:00
* not enough history pages:
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
* it could be a random read
*/
readahead: make context readahead more conservative This helps performance on moderately dense random reads on SSD. Transaction-Per-Second numbers provided by Taobao: QPS case ------------------------------------------------------- 7536 disable context readahead totally w/ patch: 7129 slower size rampup and start RA on the 3rd read 6717 slower size rampup w/o patch: 5581 unmodified context readahead Before, readahead will be started whenever reading page N+1 when it happen to read N recently. After patch, we'll only start readahead when *three* random reads happen to access pages N, N+1, N+2. The probability of this happening is extremely low for pure random reads, unless they are very dense, which actually deserves some readahead. Also start with a smaller readahead window. The impact to interleaved sequential reads should be small, because for a long run stream, the the small readahead window rampup phase is negletable. The context readahead actually benefits clustered random reads on HDD whose seek cost is pretty high. However as SSD is increasingly used for random read workloads it's better for the context readahead to concentrate on interleaved sequential reads. Another SSD rand read test from Miao # file size: 2GB # read IO amount: 625MB sysbench --test=fileio \ --max-requests=10000 \ --num-threads=1 \ --file-num=1 \ --file-block-size=64K \ --file-test-mode=rndrd \ --file-fsync-freq=0 \ --file-fsync-end=off run shows the performance of btrfs grows up from 69MB/s to 121MB/s, ext4 from 104MB/s to 121MB/s. Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Tested-by: Tao Ma <tm@tao.ma> Tested-by: Miao Xie <miaox@cn.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-12 05:21:47 +08:00
if (size <= req_size)
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
return 0;
/*
* starts from beginning of file:
* it is a strong indication of long-run stream (or whole-file-read)
*/
if (size >= offset)
size *= 2;
ra->start = offset;
readahead: make context readahead more conservative This helps performance on moderately dense random reads on SSD. Transaction-Per-Second numbers provided by Taobao: QPS case ------------------------------------------------------- 7536 disable context readahead totally w/ patch: 7129 slower size rampup and start RA on the 3rd read 6717 slower size rampup w/o patch: 5581 unmodified context readahead Before, readahead will be started whenever reading page N+1 when it happen to read N recently. After patch, we'll only start readahead when *three* random reads happen to access pages N, N+1, N+2. The probability of this happening is extremely low for pure random reads, unless they are very dense, which actually deserves some readahead. Also start with a smaller readahead window. The impact to interleaved sequential reads should be small, because for a long run stream, the the small readahead window rampup phase is negletable. The context readahead actually benefits clustered random reads on HDD whose seek cost is pretty high. However as SSD is increasingly used for random read workloads it's better for the context readahead to concentrate on interleaved sequential reads. Another SSD rand read test from Miao # file size: 2GB # read IO amount: 625MB sysbench --test=fileio \ --max-requests=10000 \ --num-threads=1 \ --file-num=1 \ --file-block-size=64K \ --file-test-mode=rndrd \ --file-fsync-freq=0 \ --file-fsync-end=off run shows the performance of btrfs grows up from 69MB/s to 121MB/s, ext4 from 104MB/s to 121MB/s. Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Tested-by: Tao Ma <tm@tao.ma> Tested-by: Miao Xie <miaox@cn.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-09-12 05:21:47 +08:00
ra->size = min(size + req_size, max);
ra->async_size = 1;
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
return 1;
}
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
/*
* A minimal readahead algorithm for trivial sequential/random reads.
*/
static unsigned long
ondemand_readahead(struct address_space *mapping,
struct file_ra_state *ra, struct file *filp,
bool hit_readahead_marker, pgoff_t offset,
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
unsigned long req_size)
{
struct backing_dev_info *bdi = inode_to_bdi(mapping->host);
unsigned long max_pages = ra->ra_pages;
unsigned long add_pages;
2013-11-13 07:08:16 +08:00
pgoff_t prev_offset;
/*
* If the request exceeds the readahead window, allow the read to
* be up to the optimal hardware IO size
*/
if (req_size > max_pages && bdi->io_pages > max_pages)
max_pages = min(req_size, bdi->io_pages);
/*
* start of file
*/
if (!offset)
goto initial_readahead;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
/*
* It's the expected callback offset, assume sequential access.
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
* Ramp up sizes, and push forward the readahead window.
*/
if ((offset == (ra->start + ra->size - ra->async_size) ||
offset == (ra->start + ra->size))) {
ra->start += ra->size;
ra->size = get_next_ra_size(ra, max_pages);
ra->async_size = ra->size;
goto readit;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
}
readahead: basic support of interleaved reads This is a simplified version of the pagecache context based readahead. It handles the case of multiple threads reading on the same fd and invalidating each others' readahead state. It does the trick by scanning the pagecache and recovering the current read stream's readahead status. The algorithm works in a opportunistic way, in that it does not try to detect interleaved reads _actively_, which requires a probe into the page cache (which means a little more overhead for random reads). It only tries to handle a previously started sequential readahead whose state was overwritten by another concurrent stream, and it can do this job pretty well. Negative and positive examples(or what you can expect from it): 1) it cannot detect and serve perfect request-by-request interleaved reads right: time stream 1 stream 2 0 1 1 1001 2 2 3 1002 4 3 5 1003 6 4 7 1004 8 5 9 1005 Here no single readahead will be carried out. 2) However, if it's two concurrent reads by two threads, the chance of the initial sequential readahead be started is huge. Once the first sequential readahead is started for a stream, this patch will ensure that the readahead window continues to rampup and won't be disturbed by other streams. time stream 1 stream 2 0 1 1 2 2 1001 3 3 4 1002 5 1003 6 4 7 5 8 1004 9 6 10 1005 11 7 12 1006 13 1007 Here stream 1 will start a readahead at page 2, and stream 2 will start its first readahead at page 1003. From then on the two streams will be served right. Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 16:24:34 +08:00
/*
* Hit a marked page without valid readahead state.
* E.g. interleaved reads.
* Query the pagecache for async_size, which normally equals to
* readahead size. Ramp it up and use it as the new readahead size.
*/
if (hit_readahead_marker) {
pgoff_t start;
rcu_read_lock();
start = page_cache_next_miss(mapping, offset + 1, max_pages);
rcu_read_unlock();
readahead: basic support of interleaved reads This is a simplified version of the pagecache context based readahead. It handles the case of multiple threads reading on the same fd and invalidating each others' readahead state. It does the trick by scanning the pagecache and recovering the current read stream's readahead status. The algorithm works in a opportunistic way, in that it does not try to detect interleaved reads _actively_, which requires a probe into the page cache (which means a little more overhead for random reads). It only tries to handle a previously started sequential readahead whose state was overwritten by another concurrent stream, and it can do this job pretty well. Negative and positive examples(or what you can expect from it): 1) it cannot detect and serve perfect request-by-request interleaved reads right: time stream 1 stream 2 0 1 1 1001 2 2 3 1002 4 3 5 1003 6 4 7 1004 8 5 9 1005 Here no single readahead will be carried out. 2) However, if it's two concurrent reads by two threads, the chance of the initial sequential readahead be started is huge. Once the first sequential readahead is started for a stream, this patch will ensure that the readahead window continues to rampup and won't be disturbed by other streams. time stream 1 stream 2 0 1 1 2 2 1001 3 3 4 1002 5 1003 6 4 7 5 8 1004 9 6 10 1005 11 7 12 1006 13 1007 Here stream 1 will start a readahead at page 2, and stream 2 will start its first readahead at page 1003. From then on the two streams will be served right. Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 16:24:34 +08:00
if (!start || start - offset > max_pages)
readahead: basic support of interleaved reads This is a simplified version of the pagecache context based readahead. It handles the case of multiple threads reading on the same fd and invalidating each others' readahead state. It does the trick by scanning the pagecache and recovering the current read stream's readahead status. The algorithm works in a opportunistic way, in that it does not try to detect interleaved reads _actively_, which requires a probe into the page cache (which means a little more overhead for random reads). It only tries to handle a previously started sequential readahead whose state was overwritten by another concurrent stream, and it can do this job pretty well. Negative and positive examples(or what you can expect from it): 1) it cannot detect and serve perfect request-by-request interleaved reads right: time stream 1 stream 2 0 1 1 1001 2 2 3 1002 4 3 5 1003 6 4 7 1004 8 5 9 1005 Here no single readahead will be carried out. 2) However, if it's two concurrent reads by two threads, the chance of the initial sequential readahead be started is huge. Once the first sequential readahead is started for a stream, this patch will ensure that the readahead window continues to rampup and won't be disturbed by other streams. time stream 1 stream 2 0 1 1 2 2 1001 3 3 4 1002 5 1003 6 4 7 5 8 1004 9 6 10 1005 11 7 12 1006 13 1007 Here stream 1 will start a readahead at page 2, and stream 2 will start its first readahead at page 1003. From then on the two streams will be served right. Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 16:24:34 +08:00
return 0;
ra->start = start;
ra->size = start - offset; /* old async_size */
ra->size += req_size;
ra->size = get_next_ra_size(ra, max_pages);
readahead: basic support of interleaved reads This is a simplified version of the pagecache context based readahead. It handles the case of multiple threads reading on the same fd and invalidating each others' readahead state. It does the trick by scanning the pagecache and recovering the current read stream's readahead status. The algorithm works in a opportunistic way, in that it does not try to detect interleaved reads _actively_, which requires a probe into the page cache (which means a little more overhead for random reads). It only tries to handle a previously started sequential readahead whose state was overwritten by another concurrent stream, and it can do this job pretty well. Negative and positive examples(or what you can expect from it): 1) it cannot detect and serve perfect request-by-request interleaved reads right: time stream 1 stream 2 0 1 1 1001 2 2 3 1002 4 3 5 1003 6 4 7 1004 8 5 9 1005 Here no single readahead will be carried out. 2) However, if it's two concurrent reads by two threads, the chance of the initial sequential readahead be started is huge. Once the first sequential readahead is started for a stream, this patch will ensure that the readahead window continues to rampup and won't be disturbed by other streams. time stream 1 stream 2 0 1 1 2 2 1001 3 3 4 1002 5 1003 6 4 7 5 8 1004 9 6 10 1005 11 7 12 1006 13 1007 Here stream 1 will start a readahead at page 2, and stream 2 will start its first readahead at page 1003. From then on the two streams will be served right. Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 16:24:34 +08:00
ra->async_size = ra->size;
goto readit;
}
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
/*
* oversize read
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
*/
if (req_size > max_pages)
goto initial_readahead;
/*
* sequential cache miss
2013-11-13 07:08:16 +08:00
* trivial case: (offset - prev_offset) == 1
* unaligned reads: (offset - prev_offset) == 0
*/
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
prev_offset = (unsigned long long)ra->prev_pos >> PAGE_SHIFT;
2013-11-13 07:08:16 +08:00
if (offset - prev_offset <= 1UL)
goto initial_readahead;
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
/*
* Query the page cache and look for the traces(cached history pages)
* that a sequential stream would leave behind.
*/
if (try_context_readahead(mapping, ra, offset, req_size, max_pages))
readahead: introduce context readahead algorithm Introduce page cache context based readahead algorithm. This is to better support concurrent read streams in general. RATIONALE --------- The current readahead algorithm detects interleaved reads in a _passive_ way. Given a sequence of interleaved streams 1,1001,2,1002,3,4,1003,5,1004,1005,6,... By checking for (offset == prev_offset + 1), it will discover the sequentialness between 3,4 and between 1004,1005, and start doing sequential readahead for the individual streams since page 4 and page 1005. The context readahead algorithm guarantees to discover the sequentialness no matter how the streams are interleaved. For the above example, it will start sequential readahead since page 2 and 1002. The trick is to poke for page @offset-1 in the page cache when it has no other clues on the sequentialness of request @offset: if the current requenst belongs to a sequential stream, that stream must have accessed page @offset-1 recently, and the page will still be cached now. So if page @offset-1 is there, we can take request @offset as a sequential access. BENEFICIARIES ------------- - strictly interleaved reads i.e. 1,1001,2,1002,3,1003,... the current readahead will take them as silly random reads; the context readahead will take them as two sequential streams. - cooperative IO processes i.e. NFS and SCST They create a thread pool, farming off (sequential) IO requests to different threads which will be performing interleaved IO. It was not easy(or possible) to reliably tell from file->f_ra all those cooperative processes working on the same sequential stream, since they will have different file->f_ra instances. And NFSD's file->f_ra is particularly unusable, since their file objects are dynamically created for each request. The nfsd does have code trying to restore the f_ra bits, but not satisfactory. The new scheme is to detect the sequential pattern via looking up the page cache, which provides one single and consistent view of the pages recently accessed. That makes sequential detection for cooperative processes possible. USER REPORT ----------- Vladislav recommends the addition of context readahead as a result of his SCST benchmarks. It leads to 6%~40% performance gains in various cases and achieves equal performance in others. http://lkml.org/lkml/2009/3/19/239 OVERHEADS --------- In theory, it introduces one extra page cache lookup per random read. However the below benchmark shows context readahead to be slightly faster, wondering.. Randomly reading 200MB amount of data on a sparse file, repeat 20 times for each block size. The average throughputs are: original ra context ra gain 4K random reads: 65.561MB/s 65.648MB/s +0.1% 16K random reads: 124.767MB/s 124.951MB/s +0.1% 64K random reads: 162.123MB/s 162.278MB/s +0.1% Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Jeff Moyer <jmoyer@redhat.com> Tested-by: Vladislav Bolkhovitin <vst@vlnb.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-06-17 06:31:36 +08:00
goto readit;
/*
* standalone, small random read
* Read as is, and do not pollute the readahead state.
*/
return __do_page_cache_readahead(mapping, filp, offset, req_size, 0);
initial_readahead:
ra->start = offset;
ra->size = get_init_ra_size(req_size, max_pages);
ra->async_size = ra->size > req_size ? ra->size - req_size : ra->size;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
readit:
/*
* Will this read hit the readahead marker made by itself?
* If so, trigger the readahead marker hit now, and merge
* the resulted next readahead window into the current one.
* Take care of maximum IO pages as above.
*/
if (offset == ra->start && ra->size == ra->async_size) {
add_pages = get_next_ra_size(ra, max_pages);
if (ra->size + add_pages <= max_pages) {
ra->async_size = add_pages;
ra->size += add_pages;
} else {
ra->size = max_pages;
ra->async_size = max_pages >> 1;
}
}
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
return ra_submit(ra, mapping, filp);
}
/**
* page_cache_sync_readahead - generic file readahead
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
* @mapping: address_space which holds the pagecache and I/O vectors
* @ra: file_ra_state which holds the readahead state
* @filp: passed on to ->readpage() and ->readpages()
* @offset: start offset into @mapping, in pagecache page-sized units
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
* @req_size: hint: total size of the read which the caller is performing in
* pagecache pages
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
*
* page_cache_sync_readahead() should be called when a cache miss happened:
* it will submit the read. The readahead logic may decide to piggyback more
* pages onto the read request if access patterns suggest it will improve
* performance.
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
*/
void page_cache_sync_readahead(struct address_space *mapping,
struct file_ra_state *ra, struct file *filp,
pgoff_t offset, unsigned long req_size)
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
{
/* no read-ahead */
if (!ra->ra_pages)
return;
if (blk_cgroup_congested())
return;
readahead: introduce FMODE_RANDOM for POSIX_FADV_RANDOM This fixes inefficient page-by-page reads on POSIX_FADV_RANDOM. POSIX_FADV_RANDOM used to set ra_pages=0, which leads to poor performance: a 16K read will be carried out in 4 _sync_ 1-page reads. In other places, ra_pages==0 means - it's ramfs/tmpfs/hugetlbfs/sysfs/configfs - some IO error happened where multi-page read IO won't help or should be avoided. POSIX_FADV_RANDOM actually want a different semantics: to disable the *heuristic* readahead algorithm, and to use a dumb one which faithfully submit read IO for whatever application requests. So introduce a flag FMODE_RANDOM for POSIX_FADV_RANDOM. Note that the random hint is not likely to help random reads performance noticeably. And it may be too permissive on huge request size (its IO size is not limited by read_ahead_kb). In Quentin's report (http://lkml.org/lkml/2009/12/24/145), the overall (NFS read) performance of the application increased by 313%! Tested-by: Quentin Barnes <qbarnes+nfs@yahoo-inc.com> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Andi Kleen <andi@firstfloor.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: David Howells <dhowells@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@infradead.org> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: <stable@kernel.org> [2.6.33.x] Cc: <qbarnes+nfs@yahoo-inc.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-06 05:42:03 +08:00
/* be dumb */
if (filp && (filp->f_mode & FMODE_RANDOM)) {
readahead: introduce FMODE_RANDOM for POSIX_FADV_RANDOM This fixes inefficient page-by-page reads on POSIX_FADV_RANDOM. POSIX_FADV_RANDOM used to set ra_pages=0, which leads to poor performance: a 16K read will be carried out in 4 _sync_ 1-page reads. In other places, ra_pages==0 means - it's ramfs/tmpfs/hugetlbfs/sysfs/configfs - some IO error happened where multi-page read IO won't help or should be avoided. POSIX_FADV_RANDOM actually want a different semantics: to disable the *heuristic* readahead algorithm, and to use a dumb one which faithfully submit read IO for whatever application requests. So introduce a flag FMODE_RANDOM for POSIX_FADV_RANDOM. Note that the random hint is not likely to help random reads performance noticeably. And it may be too permissive on huge request size (its IO size is not limited by read_ahead_kb). In Quentin's report (http://lkml.org/lkml/2009/12/24/145), the overall (NFS read) performance of the application increased by 313%! Tested-by: Quentin Barnes <qbarnes+nfs@yahoo-inc.com> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Cc: Nick Piggin <npiggin@suse.de> Cc: Andi Kleen <andi@firstfloor.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: David Howells <dhowells@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@infradead.org> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: Chuck Lever <chuck.lever@oracle.com> Cc: <stable@kernel.org> [2.6.33.x] Cc: <qbarnes+nfs@yahoo-inc.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2010-03-06 05:42:03 +08:00
force_page_cache_readahead(mapping, filp, offset, req_size);
return;
}
/* do read-ahead */
ondemand_readahead(mapping, ra, filp, false, offset, req_size);
}
EXPORT_SYMBOL_GPL(page_cache_sync_readahead);
/**
* page_cache_async_readahead - file readahead for marked pages
* @mapping: address_space which holds the pagecache and I/O vectors
* @ra: file_ra_state which holds the readahead state
* @filp: passed on to ->readpage() and ->readpages()
* @page: the page at @offset which has the PG_readahead flag set
* @offset: start offset into @mapping, in pagecache page-sized units
* @req_size: hint: total size of the read which the caller is performing in
* pagecache pages
*
* page_cache_async_readahead() should be called when a page is used which
* has the PG_readahead flag; this is a marker to suggest that the application
* has used up enough of the readahead window that we should start pulling in
* more pages.
*/
void
page_cache_async_readahead(struct address_space *mapping,
struct file_ra_state *ra, struct file *filp,
struct page *page, pgoff_t offset,
unsigned long req_size)
{
/* no read-ahead */
if (!ra->ra_pages)
return;
/*
* Same bit is used for PG_readahead and PG_reclaim.
*/
if (PageWriteback(page))
return;
ClearPageReadahead(page);
/*
* Defer asynchronous read-ahead on IO congestion.
*/
if (inode_read_congested(mapping->host))
return;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
if (blk_cgroup_congested())
return;
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
/* do read-ahead */
ondemand_readahead(mapping, ra, filp, true, offset, req_size);
readahead: on-demand readahead logic This is a minimal readahead algorithm that aims to replace the current one. It is more flexible and reliable, while maintaining almost the same behavior and performance. Also it is full integrated with adaptive readahead. It is designed to be called on demand: - on a missing page, to do synchronous readahead - on a lookahead page, to do asynchronous readahead In this way it eliminated the awkward workarounds for cache hit/miss, readahead thrashing, retried read, and unaligned read. It also adopts the data structure introduced by adaptive readahead, parameterizes readahead pipelining with `lookahead_index', and reduces the current/ahead windows to one single window. HEURISTICS The logic deals with four cases: - sequential-next found a consistent readahead window, so push it forward - random standalone small read, so read as is - sequential-first create a new readahead window for a sequential/oversize request - lookahead-clueless hit a lookahead page not associated with the readahead window, so create a new readahead window and ramp it up In each case, three parameters are determined: - readahead index: where the next readahead begins - readahead size: how much to readahead - lookahead size: when to do the next readahead (for pipelining) BEHAVIORS The old behaviors are maximally preserved for trivial sequential/random reads. Notable changes are: - It no longer imposes strict sequential checks. It might help some interleaved cases, and clustered random reads. It does introduce risks of a random lookahead hit triggering an unexpected readahead. But in general it is more likely to do good than to do evil. - Interleaved reads are supported in a minimal way. Their chances of being detected and proper handled are still low. - Readahead thrashings are better handled. The current readahead leads to tiny average I/O sizes, because it never turn back for the thrashed pages. They have to be fault in by do_generic_mapping_read() one by one. Whereas the on-demand readahead will redo readahead for them. OVERHEADS The new code reduced the overheads of - excessively calling the readahead routine on small sized reads (the current readahead code insists on seeing all requests) - doing a lot of pointless page-cache lookups for small cached files (the current readahead only turns itself off after 256 cache hits, unfortunately most files are < 1MB, so never see that chance) That accounts for speedup of - 0.3% on 1-page sequential reads on sparse file - 1.2% on 1-page cache hot sequential reads - 3.2% on 256-page cache hot sequential reads - 1.3% on cache hot `tar /lib` However, it does introduce one extra page-cache lookup per cache miss, which impacts random reads slightly. That's 1% overheads for 1-page random reads on sparse file. PERFORMANCE The basic benchmark setup is - 2.6.20 kernel with on-demand readahead - 1MB max readahead size - 2.9GHz Intel Core 2 CPU - 2GB memory - 160G/8M Hitachi SATA II 7200 RPM disk The benchmarks show that - it maintains the same performance for trivial sequential/random reads - sysbench/OLTP performance on MySQL gains up to 8% - performance on readahead thrashing gains up to 3 times iozone throughput (KB/s): roughly the same ========================================== iozone -c -t1 -s 4096m -r 64k 2.6.20 on-demand gain first run " Initial write " 61437.27 64521.53 +5.0% " Rewrite " 47893.02 48335.20 +0.9% " Read " 62111.84 62141.49 +0.0% " Re-read " 62242.66 62193.17 -0.1% " Reverse Read " 50031.46 49989.79 -0.1% " Stride read " 8657.61 8652.81 -0.1% " Random read " 13914.28 13898.23 -0.1% " Mixed workload " 19069.27 19033.32 -0.2% " Random write " 14849.80 14104.38 -5.0% " Pwrite " 62955.30 65701.57 +4.4% " Pread " 62209.99 62256.26 +0.1% second run " Initial write " 60810.31 66258.69 +9.0% " Rewrite " 49373.89 57833.66 +17.1% " Read " 62059.39 62251.28 +0.3% " Re-read " 62264.32 62256.82 -0.0% " Reverse Read " 49970.96 50565.72 +1.2% " Stride read " 8654.81 8638.45 -0.2% " Random read " 13901.44 13949.91 +0.3% " Mixed workload " 19041.32 19092.04 +0.3% " Random write " 14019.99 14161.72 +1.0% " Pwrite " 64121.67 68224.17 +6.4% " Pread " 62225.08 62274.28 +0.1% In summary, writes are unstable, reads are pretty close on average: access pattern 2.6.20 on-demand gain Read 62085.61 62196.38 +0.2% Re-read 62253.49 62224.99 -0.0% Reverse Read 50001.21 50277.75 +0.6% Stride read 8656.21 8645.63 -0.1% Random read 13907.86 13924.07 +0.1% Mixed workload 19055.29 19062.68 +0.0% Pread 62217.53 62265.27 +0.1% aio-stress: roughly the same ============================ aio-stress -l -s4096 -r128 -t1 -o1 knoppix511-dvd-cn.iso aio-stress -l -s4096 -r128 -t1 -o3 knoppix511-dvd-cn.iso 2.6.20 on-demand delta sequential 92.57s 92.54s -0.0% random 311.87s 312.15s +0.1% sysbench fileio: roughly the same ================================= sysbench --test=fileio --file-io-mode=async --file-test-mode=rndrw \ --file-total-size=4G --file-block-size=64K \ --num-threads=001 --max-requests=10000 --max-time=900 run threads 2.6.20 on-demand delta first run 1 59.1974s 59.2262s +0.0% 2 58.0575s 58.2269s +0.3% 4 48.0545s 47.1164s -2.0% 8 41.0684s 41.2229s +0.4% 16 35.8817s 36.4448s +1.6% 32 32.6614s 32.8240s +0.5% 64 23.7601s 24.1481s +1.6% 128 24.3719s 23.8225s -2.3% 256 23.2366s 22.0488s -5.1% second run 1 59.6720s 59.5671s -0.2% 8 41.5158s 41.9541s +1.1% 64 25.0200s 23.9634s -4.2% 256 22.5491s 20.9486s -7.1% Note that the numbers are not very stable because of the writes. The overall performance is close when we sum all seconds up: sum all up 495.046s 491.514s -0.7% sysbench oltp (trans/sec): up to 8% gain ======================================== sysbench --test=oltp --oltp-table-size=10000000 --oltp-read-only \ --mysql-socket=/var/run/mysqld/mysqld.sock \ --mysql-user=root --mysql-password=readahead \ --num-threads=064 --max-requests=10000 --max-time=900 run 10000-transactions run threads 2.6.20 on-demand gain 1 62.81 64.56 +2.8% 2 67.97 70.93 +4.4% 4 81.81 85.87 +5.0% 8 94.60 97.89 +3.5% 16 99.07 104.68 +5.7% 32 95.93 104.28 +8.7% 64 96.48 103.68 +7.5% 5000-transactions run 1 48.21 48.65 +0.9% 8 68.60 70.19 +2.3% 64 70.57 74.72 +5.9% 2000-transactions run 1 37.57 38.04 +1.3% 2 38.43 38.99 +1.5% 4 45.39 46.45 +2.3% 8 51.64 52.36 +1.4% 16 54.39 55.18 +1.5% 32 52.13 54.49 +4.5% 64 54.13 54.61 +0.9% That's interesting results. Some investigations show that - MySQL is accessing the db file non-uniformly: some parts are more hot than others - It is mostly doing 4-page random reads, and sometimes doing two reads in a row, the latter one triggers a 16-page readahead. - The on-demand readahead leaves many lookahead pages (flagged PG_readahead) there. Many of them will be hit, and trigger more readahead pages. Which might save more seeks. - Naturally, the readahead windows tend to lie in hot areas, and the lookahead pages in hot areas is more likely to be hit. - The more overall read density, the more possible gain. That also explains the adaptive readahead tricks for clustered random reads. readahead thrashing: 3 times better =================================== We boot kernel with "mem=128m single", and start a 100KB/s stream on every second, until reaching 200 streams. max throughput min avg I/O size 2.6.20: 5MB/s 16KB on-demand: 15MB/s 140KB Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Steven Pratt <slpratt@austin.ibm.com> Cc: Ram Pai <linuxram@us.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:48:01 +08:00
}
EXPORT_SYMBOL_GPL(page_cache_async_readahead);
ssize_t ksys_readahead(int fd, loff_t offset, size_t count)
{
ssize_t ret;
struct fd f;
ret = -EBADF;
f = fdget(fd);
if (!f.file || !(f.file->f_mode & FMODE_READ))
goto out;
/*
* The readahead() syscall is intended to run only on files
* that can execute readahead. If readahead is not possible
* on this file, then we must return -EINVAL.
*/
ret = -EINVAL;
if (!f.file->f_mapping || !f.file->f_mapping->a_ops ||
!S_ISREG(file_inode(f.file)->i_mode))
goto out;
ret = vfs_fadvise(f.file, offset, count, POSIX_FADV_WILLNEED);
out:
fdput(f);
return ret;
}
SYSCALL_DEFINE3(readahead, int, fd, loff_t, offset, size_t, count)
{
return ksys_readahead(fd, offset, count);
}