It looks like the following patch can make FQ very precise, even in VM
or stressed hosts. It matters at high pacing rates.
We take into account the difference between the time that was programmed
when last packet was sent, and current time (a drift of tens of usecs is
often observed)
Add an EWMA of the unthrottle latency to help diagnostics.
This latency is the difference between current time and oldest packet in
delayed RB-tree. This accounts for the high resolution timer latency,
but can be different under stress, as fq_check_throttled() can be
opportunistically be called from a dequeue() called after an enqueue()
for a different flow.
Tested:
// Start a 10Gbit flow
$ netperf --google-pacing-rate 1250000000 -H lpaa24 -l 10000 -- -K bbr &
Before patch :
$ sar -n DEV 10 5 | grep eth0 | grep Average
Average: eth0 17106.04 756876.84 1102.75 1119049.02 0.00 0.00 0.52
After patch :
$ sar -n DEV 10 5 | grep eth0 | grep Average
Average: eth0 17867.00 800245.90 1151.77 1183172.12 0.00 0.00 0.52
A new iproute2 tc can output the 'unthrottle latency' :
$ tc -s qd sh dev eth0 | grep latency
0 gc, 0 highprio, 32490767 throttled, 2382 ns latency
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit adds to the fq module a low_rate_threshold parameter to
insert a delay after all packets if the socket requests a pacing rate
below the threshold.
This helps achieve more precise control of the sending rate with
low-rate paths, especially policers. The basic issue is that if a
congestion control module detects a policer at a certain rate, it may
want fq to be able to shape to that policed rate. That way the sender
can avoid policer drops by having the packets arrive at the policer at
or just under the policed rate.
The default threshold of 550Kbps was chosen analytically so that for
policers or links at 500Kbps or 512Kbps fq would very likely invoke
this mechanism, even if the pacing rate was briefly slightly above the
available bandwidth. This value was then empirically validated with
two years of production testing on YouTube video servers.
Signed-off-by: Van Jacobson <vanj@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Nandita Dukkipati <nanditad@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On small embedded routers, one wants to control maximal amount of
memory used by fq_codel, instead of controlling number of packets or
bytes, since GRO/TSO make these not practical.
Assuming skb->truesize is accurate, we have to keep track of
skb->truesize sum for skbs in queue.
This patch adds a new TCA_FQ_CODEL_MEMORY_LIMIT attribute.
I chose a default value of 32 MBytes, which looks reasonable even
for heavy duty usages. (Prior fq_codel users should not be hurt
when they upgrade their kernels)
Two fields are added to tc_fq_codel_qd_stats to report :
- Current memory usage
- Number of drops caused by memory limits
# tc qd replace dev eth1 root est 1sec 4sec fq_codel memory_limit 4M
..
# tc -s -d qd sh dev eth1
qdisc fq_codel 8008: root refcnt 257 limit 10240p flows 1024
quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
Sent 2083566791363 bytes 1376214889 pkt (dropped 4994406, overlimits 0
requeues 21705223)
rate 9841Mbit 812549pps backlog 3906120b 376p requeues 21705223
maxpacket 68130 drop_overlimit 4994406 new_flow_count 28855414
ecn_mark 0 memory_used 4190048 drop_overmemory 4994406
new_flows_len 1 old_flows_len 177
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Dave Täht <dave.taht@gmail.com>
Cc: Sebastian Möller <moeller0@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
In presence of inelastic flows and stress, we can call
fq_codel_drop() for every packet entering fq_codel qdisc.
fq_codel_drop() is quite expensive, as it does a linear scan
of 4 KB of memory to find a fat flow.
Once found, it drops the oldest packet of this flow.
Instead of dropping a single packet, try to drop 50% of the backlog
of this fat flow, with a configurable limit of 64 packets per round.
TCA_FQ_CODEL_DROP_BATCH_SIZE is the new attribute to make this
limit configurable.
With this strategy the 4 KB search is amortized to a single cache line
per drop [1], so fq_codel_drop() no longer appears at the top of kernel
profile in presence of few inelastic flows.
[1] Assuming a 64byte cache line, and 1024 buckets
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dave Taht <dave.taht@gmail.com>
Cc: Jonathan Morton <chromatix99@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Dave Taht
Signed-off-by: David S. Miller <davem@davemloft.net>
This work adds a generalization of the ingress qdisc as a qdisc holding
only classifiers. The clsact qdisc works on ingress, but also on egress.
In both cases, it's execution happens without taking the qdisc lock, and
the main difference for the egress part compared to prior version of [1]
is that this can be applied with _any_ underlying real egress qdisc (also
classless ones).
Besides solving the use-case of [1], that is, allowing for more programmability
on assigning skb->priority for the mqprio case that is supported by most
popular 10G+ NICs, it also opens up a lot more flexibility for other tc
applications. The main work on classification can already be done at clsact
egress time if the use-case allows and state stored for later retrieval
f.e. again in skb->priority with major/minors (which is checked by most
classful qdiscs before consulting tc_classify()) and/or in other skb fields
like skb->tc_index for some light-weight post-processing to get to the
eventual classid in case of a classful qdisc. Another use case is that
the clsact egress part allows to have a central egress counterpart to
the ingress classifiers, so that classifiers can easily share state (e.g.
in cls_bpf via eBPF maps) for ingress and egress.
Currently, default setups like mq + pfifo_fast would require for this to
use, for example, prio qdisc instead (to get a tc_classify() run) and to
duplicate the egress classifier for each queue. With clsact, it allows
for leaving the setup as is, it can additionally assign skb->priority to
put the skb in one of pfifo_fast's bands and it can share state with maps.
Moreover, we can access the skb's dst entry (f.e. to retrieve tclassid)
w/o the need to perform a skb_dst_force() to hold on to it any longer. In
lwt case, we can also use this facility to setup dst metadata via cls_bpf
(bpf_skb_set_tunnel_key()) without needing a real egress qdisc just for
that (case of IFF_NO_QUEUE devices, for example).
The realization can be done without any changes to the scheduler core
framework. All it takes is that we have two a-priori defined minors/child
classes, where we can mux between ingress and egress classifier list
(dev->ingress_cl_list and dev->egress_cl_list, latter stored close to
dev->_tx to avoid extra cacheline miss for moderate loads). The egress
part is a bit similar modelled to handle_ing() and patched to a noop in
case the functionality is not used. Both handlers are now called
sch_handle_ingress() and sch_handle_egress(), code sharing among the two
doesn't seem practical as there are various minor differences in both
paths, so that making them conditional in a single handler would rather
slow things down.
Full compatibility to ingress qdisc is provided as well. Since both
piggyback on TC_H_CLSACT, only one of them (ingress/clsact) can exist
per netdevice, and thus ingress qdisc specific behaviour can be retained
for user space. This means, either a user does 'tc qdisc add dev foo ingress'
and configures ingress qdisc as usual, or the 'tc qdisc add dev foo clsact'
alternative, where both, ingress and egress classifier can be configured
as in the below example. ingress qdisc supports attaching classifier to any
minor number whereas clsact has two fixed minors for muxing between the
lists, therefore to not break user space setups, they are better done as
two separate qdiscs.
I decided to extend the sch_ingress module with clsact functionality so
that commonly used code can be reused, the module is being aliased with
sch_clsact so that it can be auto-loaded properly. Alternative would have been
to add a flag when initializing ingress to alter its behaviour plus aliasing
to a different name (as it's more than just ingress). However, the first would
end up, based on the flag, choosing the new/old behaviour by calling different
function implementations to handle each anyway, the latter would require to
register ingress qdisc once again under different alias. So, this really begs
to provide a minimal, cleaner approach to have Qdisc_ops and Qdisc_class_ops
by its own that share callbacks used by both.
Example, adding qdisc:
# tc qdisc add dev foo clsact
# tc qdisc show dev foo
qdisc mq 0: root
qdisc pfifo_fast 0: parent :1 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
qdisc pfifo_fast 0: parent :2 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
qdisc pfifo_fast 0: parent :3 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
qdisc pfifo_fast 0: parent :4 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
qdisc clsact ffff: parent ffff:fff1
Adding filters (deleting, etc works analogous by specifying ingress/egress):
# tc filter add dev foo ingress bpf da obj bar.o sec ingress
# tc filter add dev foo egress bpf da obj bar.o sec egress
# tc filter show dev foo ingress
filter protocol all pref 49152 bpf
filter protocol all pref 49152 bpf handle 0x1 bar.o:[ingress] direct-action
# tc filter show dev foo egress
filter protocol all pref 49152 bpf
filter protocol all pref 49152 bpf handle 0x1 bar.o:[egress] direct-action
A 'tc filter show dev foo' or 'tc filter show dev foo parent ffff:' will
show an empty list for clsact. Either using the parent names (ingress/egress)
or specifying the full major/minor will then show the related filter lists.
Prior work on a mqprio prequeue() facility [1] was done mainly by John Fastabend.
[1] http://patchwork.ozlabs.org/patch/512949/
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In a GRED qdisc, if the default "virtual queue" (VQ) does not have drop
parameters configured, then packets for the default VQ are not subjected
to RED and are only dropped if the queue is larger than the net_device's
tx_queue_len. This behavior is useful for WRED mode, since these packets
will still influence the calculated average queue length and (therefore)
the drop probability for all of the other VQs. However, for some drivers
tx_queue_len is zero. In other cases the user may wish to make the limit
the same for all VQs (including the default VQ with no drop parameters).
This change adds a TCA_GRED_LIMIT attribute to set the GRED queue limit,
in bytes, during qdisc setup. (This limit is in bytes to be consistent
with the drop parameters.) The default limit is the same as for a bfifo
queue (tx_queue_len * psched_mtu). If the drop parameters of any VQ are
configured with a smaller limit than the GRED queue limit, that VQ will
still observe the smaller limit instead.
Signed-off-by: David Ward <david.ward@ll.mit.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
For DCTCP or similar ECN based deployments on fabrics with shallow
buffers, hosts are responsible for a good part of the buffering.
This patch adds an optional ce_threshold to codel & fq_codel qdiscs,
so that DCTCP can have feedback from queuing in the host.
A DCTCP enabled egress port simply have a queue occupancy threshold
above which ECT packets get CE mark.
In codel language this translates to a sojourn time, so that one doesn't
have to worry about bytes or bandwidth but delays.
This makes the host an active participant in the health of the whole
network.
This also helps experimenting DCTCP in a setup without DCTCP compliant
fabric.
On following example, ce_threshold is set to 1ms, and we can see from
'ldelay xxx us' that TCP is not trying to go around the 5ms codel
target.
Queue has more capacity to absorb inelastic bursts (say from UDP
traffic), as queues are maintained to an optimal level.
lpaa23:~# ./tc -s -d qd sh dev eth1
qdisc mq 1: dev eth1 root
Sent 87910654696 bytes 58065331 pkt (dropped 0, overlimits 0 requeues 42961)
backlog 3108242b 364p requeues 42961
qdisc codel 8063: dev eth1 parent 1:1 limit 1000p target 5.0ms ce_threshold 1.0ms interval 100.0ms
Sent 7363778701 bytes 4863809 pkt (dropped 0, overlimits 0 requeues 5503)
rate 2348Mbit 193919pps backlog 255866b 46p requeues 5503
count 0 lastcount 0 ldelay 1.0ms drop_next 0us
maxpacket 68130 ecn_mark 0 drop_overlimit 0 ce_mark 72384
qdisc codel 8064: dev eth1 parent 1:2 limit 1000p target 5.0ms ce_threshold 1.0ms interval 100.0ms
Sent 7636486190 bytes 5043942 pkt (dropped 0, overlimits 0 requeues 5186)
rate 2319Mbit 191538pps backlog 207418b 64p requeues 5186
count 0 lastcount 0 ldelay 694us drop_next 0us
maxpacket 68130 ecn_mark 0 drop_overlimit 0 ce_mark 69873
qdisc codel 8065: dev eth1 parent 1:3 limit 1000p target 5.0ms ce_threshold 1.0ms interval 100.0ms
Sent 11569360142 bytes 7641602 pkt (dropped 0, overlimits 0 requeues 5554)
rate 3041Mbit 251096pps backlog 210446b 59p requeues 5554
count 0 lastcount 0 ldelay 889us drop_next 0us
maxpacket 68130 ecn_mark 0 drop_overlimit 0 ce_mark 37780
...
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Glenn Judd <glenn.judd@morganstanley.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
FQ has a fast path for skb attached to a socket, as it does not
have to compute a flow hash. But for other packets, FQ being non
stochastic means that hosts exposed to random Internet traffic
can allocate million of flows structure (104 bytes each) pretty
easily. Not only host can OOM, but lookup in RB trees can take
too much cpu and memory resources.
This patch adds a new attribute, orphan_mask, that is adding
possibility of having a stochastic hash for orphaned skb.
Its default value is 1024 slots, to mimic SFQ behavior.
Note: This does not apply to locally generated TCP traffic,
and no locally generated traffic will share a flow structure
with another perfect or stochastic flow.
This patch also handles the specific case of SYNACK messages:
They are attached to the listener socket, and therefore all map
to a single hash bucket. If listener have set SO_MAX_PACING_RATE,
hoping to have new accepted socket inherit this rate, SYNACK
might be paced and even dropped.
This is very similar to an internal patch Google have used more
than one year.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Proportional Integral controller Enhanced (PIE) is a scheduler to address the
bufferbloat problem.
>From the IETF draft below:
" Bufferbloat is a phenomenon where excess buffers in the network cause high
latency and jitter. As more and more interactive applications (e.g. voice over
IP, real time video streaming and financial transactions) run in the Internet,
high latency and jitter degrade application performance. There is a pressing
need to design intelligent queue management schemes that can control latency and
jitter; and hence provide desirable quality of service to users.
We present here a lightweight design, PIE(Proportional Integral controller
Enhanced) that can effectively control the average queueing latency to a target
value. Simulation results, theoretical analysis and Linux testbed results have
shown that PIE can ensure low latency and achieve high link utilization under
various congestion situations. The design does not require per-packet
timestamp, so it incurs very small overhead and is simple enough to implement
in both hardware and software. "
Many thanks to Dave Taht for extensive feedback, reviews, testing and
suggestions. Thanks also to Stephen Hemminger and Eric Dumazet for reviews and
suggestions. Naeem Khademi and Dave Taht independently contributed to ECN
support.
For more information, please see technical paper about PIE in the IEEE
Conference on High Performance Switching and Routing 2013. A copy of the paper
can be found at ftp://ftpeng.cisco.com/pie/.
Please also refer to the IETF draft submission at
http://tools.ietf.org/html/draft-pan-tsvwg-pie-00
All relevant code, documents and test scripts and results can be found at
ftp://ftpeng.cisco.com/pie/.
For problems with the iproute2/tc or Linux kernel code, please contact Vijay
Subramanian (vijaynsu@cisco.com or subramanian.vijay@gmail.com) Mythili Prabhu
(mysuryan@cisco.com)
Signed-off-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: Mythili Prabhu <mysuryan@cisco.com>
CC: Dave Taht <dave.taht@bufferbloat.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a new attribute to support 64bit rates so that
tc can use them to break the 32bit limit.
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we set burst to 1514 with low rate in userspace,
the kernel get a value of burst that less than 1514,
which doesn't work.
Because it may make some loss when transform burst
to buffer in userspace. This makes burst lose some
bytes, when the kernel transform the buffer back to
burst.
This patch adds two new attributes to support sending
burst/mtu to kernel directly to avoid the loss.
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements the first size-based qdisc that attempts to
differentiate between small flows and heavy-hitters. The goal is to
catch the heavy-hitters and move them to a separate queue with less
priority so that bulk traffic does not affect the latency of critical
traffic. Currently "less priority" means less weight (2:1 in
particular) in a Weighted Deficit Round Robin (WDRR) scheduler.
In essence, this patch addresses the "delay-bloat" problem due to
bloated buffers. In some systems, large queues may be necessary for
obtaining CPU efficiency, or due to the presence of unresponsive
traffic like UDP, or just a large number of connections with each
having a small amount of outstanding traffic. In these circumstances,
HHF aims to reduce the HoL blocking for latency sensitive traffic,
while not impacting the queues built up by bulk traffic. HHF can also
be used in conjunction with other AQM mechanisms such as CoDel.
To capture heavy-hitters, we implement the "multi-stage filter" design
in the following paper:
C. Estan and G. Varghese, "New Directions in Traffic Measurement and
Accounting", in ACM SIGCOMM, 2002.
Some configurable qdisc settings through 'tc':
- hhf_reset_timeout: period to reset counter values in the multi-stage
filter (default 40ms)
- hhf_admit_bytes: threshold to classify heavy-hitters
(default 128KB)
- hhf_evict_timeout: threshold to evict idle heavy-hitters
(default 1s)
- hhf_non_hh_weight: Weighted Deficit Round Robin (WDRR) weight for
non-heavy-hitters (default 2)
- hh_flows_limit: max number of heavy-hitter flow entries
(default 2048)
Note that the ratio between hhf_admit_bytes and hhf_reset_timeout
reflects the bandwidth of heavy-hitters that we attempt to capture
(25Mbps with the above default settings).
The false negative rate (heavy-hitter flows getting away unclassified)
is zero by the design of the multi-stage filter algorithm.
With 100 heavy-hitter flows, using four hashes and 4000 counters yields
a false positive rate (non-heavy-hitters mistakenly classified as
heavy-hitters) of less than 1e-4.
Signed-off-by: Terry Lam <vtlam@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For performance reasons, sch_fq tried hard to not setup timers for every
sent packet, using a quantum based heuristic : A delay is setup only if
the flow exhausted its credit.
Problem is that application limited flows can refill their credit
for every queued packet, and they can evade pacing.
This problem can also be triggered when TCP flows use small MSS values,
as TSO auto sizing builds packets that are smaller than the default fq
quantum (3028 bytes)
This patch adds a 40 ms delay to guard flow credit refill.
Fixes: afe4fd0624 ("pkt_sched: fq: Fair Queue packet scheduler")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 7eec4174ff ("pkt_sched: fq: fix non TCP flows pacing")
obsoleted TCA_FQ_FLOW_DEFAULT_RATE without notice for the users.
Suggested by David Miller
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With psched_ratecfg_precompute(), tbf can deal with 64bit rates.
Add two new attributes so that tc can use them to break the 32bit
limit.
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Suggested-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
HTB already can deal with 64bit rates, we only have to add two new
attributes so that tc can use them to break the current 32bit ABI
barrier.
TCA_HTB_RATE64 : class rate (in bytes per second)
TCA_HTB_CEIL64 : class ceil (in bytes per second)
This allows us to setup HTB on 40Gbps links, as 32bit limit is
actually ~34Gbps
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- Uses perfect flow match (not stochastic hash like SFQ/FQ_codel)
- Uses the new_flow/old_flow separation from FQ_codel
- New flows get an initial credit allowing IW10 without added delay.
- Special FIFO queue for high prio packets (no need for PRIO + FQ)
- Uses a hash table of RB trees to locate the flows at enqueue() time
- Smart on demand gc (at enqueue() time, RB tree lookup evicts old
unused flows)
- Dynamic memory allocations.
- Designed to allow millions of concurrent flows per Qdisc.
- Small memory footprint : ~8K per Qdisc, and 104 bytes per flow.
- Single high resolution timer for throttled flows (if any).
- One RB tree to link throttled flows.
- Ability to have a max rate per flow. We might add a socket option
to add per socket limitation.
Attempts have been made to add TCP pacing in TCP stack, but this
seems to add complex code to an already complex stack.
TCP pacing is welcomed for flows having idle times, as the cwnd
permits TCP stack to queue a possibly large number of packets.
This removes the 'slow start after idle' choice, hitting badly
large BDP flows, and applications delivering chunks of data
as video streams.
Nicely spaced packets :
Here interface is 10Gbit, but flow bottleneck is ~20Mbit
cwin is big, yet FQ avoids the typical bursts generated by TCP
(as in netperf TCP_RR -- -r 100000,100000)
15:01:23.545279 IP A > B: . 78193:81089(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.545394 IP B > A: . ack 81089 win 3668 <nop,nop,timestamp 11597985 1115>
15:01:23.546488 IP A > B: . 81089:83985(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.546565 IP B > A: . ack 83985 win 3668 <nop,nop,timestamp 11597986 1115>
15:01:23.547713 IP A > B: . 83985:86881(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.547778 IP B > A: . ack 86881 win 3668 <nop,nop,timestamp 11597987 1115>
15:01:23.548911 IP A > B: . 86881:89777(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.548949 IP B > A: . ack 89777 win 3668 <nop,nop,timestamp 11597988 1115>
15:01:23.550116 IP A > B: . 89777:92673(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.550182 IP B > A: . ack 92673 win 3668 <nop,nop,timestamp 11597989 1115>
15:01:23.551333 IP A > B: . 92673:95569(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.551406 IP B > A: . ack 95569 win 3668 <nop,nop,timestamp 11597991 1115>
15:01:23.552539 IP A > B: . 95569:98465(2896) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.552576 IP B > A: . ack 98465 win 3668 <nop,nop,timestamp 11597992 1115>
15:01:23.553756 IP A > B: . 98465:99913(1448) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.554138 IP A > B: P 99913:100001(88) ack 65248 win 3125 <nop,nop,timestamp 1115 11597805>
15:01:23.554204 IP B > A: . ack 100001 win 3668 <nop,nop,timestamp 11597993 1115>
15:01:23.554234 IP B > A: . 65248:68144(2896) ack 100001 win 3668 <nop,nop,timestamp 11597993 1115>
15:01:23.555620 IP B > A: . 68144:71040(2896) ack 100001 win 3668 <nop,nop,timestamp 11597993 1115>
15:01:23.557005 IP B > A: . 71040:73936(2896) ack 100001 win 3668 <nop,nop,timestamp 11597993 1115>
15:01:23.558390 IP B > A: . 73936:76832(2896) ack 100001 win 3668 <nop,nop,timestamp 11597993 1115>
15:01:23.559773 IP B > A: . 76832:79728(2896) ack 100001 win 3668 <nop,nop,timestamp 11597993 1115>
15:01:23.561158 IP B > A: . 79728:82624(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.562543 IP B > A: . 82624:85520(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.563928 IP B > A: . 85520:88416(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.565313 IP B > A: . 88416:91312(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.566698 IP B > A: . 91312:94208(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.568083 IP B > A: . 94208:97104(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.569467 IP B > A: . 97104:100000(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.570852 IP B > A: . 100000:102896(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.572237 IP B > A: . 102896:105792(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.573639 IP B > A: . 105792:108688(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.575024 IP B > A: . 108688:111584(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.576408 IP B > A: . 111584:114480(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
15:01:23.577793 IP B > A: . 114480:117376(2896) ack 100001 win 3668 <nop,nop,timestamp 11597994 1115>
TCP timestamps show that most packets from B were queued in the same ms
timeframe (TSval 1159799{3,4}), but FQ managed to send them right
in time to avoid a big burst.
In slow start or steady state, very few packets are throttled [1]
FQ gets a bunch of tunables as :
limit : max number of packets on whole Qdisc (default 10000)
flow_limit : max number of packets per flow (default 100)
quantum : the credit per RR round (default is 2 MTU)
initial_quantum : initial credit for new flows (default is 10 MTU)
maxrate : max per flow rate (default : unlimited)
buckets : number of RB trees (default : 1024) in hash table.
(consumes 8 bytes per bucket)
[no]pacing : disable/enable pacing (default is enable)
All of them can be changed on a live qdisc.
$ tc qd add dev eth0 root fq help
Usage: ... fq [ limit PACKETS ] [ flow_limit PACKETS ]
[ quantum BYTES ] [ initial_quantum BYTES ]
[ maxrate RATE ] [ buckets NUMBER ]
[ [no]pacing ]
$ tc -s -d qd
qdisc fq 8002: dev eth0 root refcnt 32 limit 10000p flow_limit 100p buckets 256 quantum 3028 initial_quantum 15140
Sent 216532416 bytes 148395 pkt (dropped 0, overlimits 0 requeues 14)
backlog 0b 0p requeues 14
511 flows, 511 inactive, 0 throttled
110 gc, 0 highprio, 0 retrans, 1143 throttled, 0 flows_plimit
[1] Except if initial srtt is overestimated, as if using
cached srtt in tcp metrics. We'll provide a fix for this issue.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 56b765b79 ("htb: improved accuracy at high rates")
broke the "linklayer atm" handling.
tc class add ... htb rate X ceil Y linklayer atm
The linklayer setting is implemented by modifying the rate table
which is send to the kernel. No direct parameter were
transferred to the kernel indicating the linklayer setting.
The commit 56b765b79 ("htb: improved accuracy at high rates")
removed the use of the rate table system.
To keep compatible with older iproute2 utils, this patch detects
the linklayer by parsing the rate table. It also supports future
versions of iproute2 to send this linklayer parameter to the
kernel directly. This is done by using the __reserved field in
struct tc_ratespec, to convey the choosen linklayer option, but
only using the lower 4 bits of this field.
Linklayer detection is limited to speeds below 100Mbit/s, because
at high rates the rtab is gets too inaccurate, so bad that
several fields contain the same values, this resembling the ATM
detect. Fields even start to contain "0" time to send, e.g. at
1000Mbit/s sending a 96 bytes packet cost "0", thus the rtab have
been more broken than we first realized.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
HTB uses an internal pfifo queue, which limit is not reported
to userland tools (tc), and value inherited from device tx_queue_len
at setup time.
Introduce TCA_HTB_DIRECT_QLEN attribute to allow finer control.
Remove two obsolete pr_err() calls as well.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Michael Kerrisk <mtk.manpages@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Dave Jones <davej@redhat.com>
Eric Dumazet