1219 lines
50 KiB
C++
1219 lines
50 KiB
C++
/*
|
|
* GlobalTagThrottler.actor.cpp
|
|
*
|
|
* This source file is part of the FoundationDB open source project
|
|
*
|
|
* Copyright 2013-2022 Apple Inc. and the FoundationDB project authors
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
|
|
#include "fdbclient/FDBTypes.h"
|
|
#include "fdbclient/TagThrottle.actor.h"
|
|
#include "fdbrpc/Smoother.h"
|
|
#include "fdbserver/TagThrottler.h"
|
|
|
|
#include <limits>
|
|
|
|
#include "flow/actorcompiler.h" // must be last include
|
|
|
|
// In the function names below, several terms are used repeatedly. The context-specific are defined here:
|
|
//
|
|
// Cost: Every read or write operation has an associated cost, determined by the number of bytes accessed.
|
|
// Global tag throttling quotas are specified in terms of the amount of this cost that can be consumed
|
|
// per second. In the global tag throttler, cost refers to the per second rate of cost consumption.
|
|
//
|
|
// TPS: Transactions per second. Quotas are not specified in terms of TPS, but the limits given to clients must
|
|
// be specified in terms of TPS because throttling is performed at the front end of transactions (before costs are
|
|
// known).
|
|
//
|
|
// Total: Refers to the total quota specified by clients through the global tag throttling API. The sum of the
|
|
// costs of all operations (cluster-wide) with a particular tag cannot exceed the tag's specified total quota,
|
|
// even if the cluster has no saturated processes.
|
|
//
|
|
// Desired TPS: Assuming that a tag is able to achieve its total quota, this is the TPS it would be able to perform.
|
|
//
|
|
// Reserved: Refers to the reserved quota specified by clients through the global tag throttling API. As long as the
|
|
// sum of the costs of all operations (cluster-wide) with a particular tag are not above the tag's
|
|
// specified reserved quota, the tag should not experience any throttling from the global tag throttler.
|
|
//
|
|
// Current [Cost|TPS]: Measuring the current throughput on the cluster, independent of any specified quotas.
|
|
//
|
|
// ThrottlingRatio: Based on the health of each storage server, a throttling ratio is provided,
|
|
// informing the global tag throttler what ratio of the current throughput can be maintained.
|
|
//
|
|
// Limiting [Cost|TPS]: Based on the health of storage servers, a limiting throughput may be enforced.
|
|
//
|
|
// Target [Cost|TPS]: Based on reserved, limiting, and desired throughputs, this is the target throughput
|
|
// that the global tag throttler aims to achieve (across all clients).
|
|
//
|
|
// PerClient TPS: Because the target throughput must be shared across multiple clients, and all clients must
|
|
// be given the same limits, a per-client limit is calculated based on the current and target throughputs.
|
|
|
|
class GlobalTagThrottlerImpl {
|
|
template <class K, class V>
|
|
static Optional<V> tryGet(std::unordered_map<K, V> const& m, K const& k) {
|
|
auto it = m.find(k);
|
|
if (it == m.end()) {
|
|
return {};
|
|
} else {
|
|
return it->second;
|
|
}
|
|
}
|
|
|
|
static Optional<double> getMin(Optional<double> a, Optional<double> b) {
|
|
if (a.present() && b.present()) {
|
|
return std::min(a.get(), b.get());
|
|
} else if (a.present()) {
|
|
return a;
|
|
} else {
|
|
return b;
|
|
}
|
|
}
|
|
|
|
static Optional<double> getMax(Optional<double> a, Optional<double> b) {
|
|
if (a.present() && b.present()) {
|
|
return std::max(a.get(), b.get());
|
|
} else if (a.present()) {
|
|
return a;
|
|
} else {
|
|
return b;
|
|
}
|
|
}
|
|
|
|
enum class LimitType { RESERVED, TOTAL };
|
|
enum class OpType { READ, WRITE };
|
|
|
|
class ThroughputCounters {
|
|
Smoother readCost;
|
|
Smoother writeCost;
|
|
|
|
public:
|
|
ThroughputCounters()
|
|
: readCost(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_FOLDING_TIME),
|
|
writeCost(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_FOLDING_TIME) {}
|
|
|
|
void updateCost(double newCost, OpType opType) {
|
|
if (opType == OpType::READ) {
|
|
readCost.setTotal(newCost);
|
|
} else {
|
|
writeCost.setTotal(newCost);
|
|
}
|
|
}
|
|
|
|
double getCost() const { return readCost.smoothTotal() + writeCost.smoothTotal(); }
|
|
};
|
|
|
|
// Track various statistics per tag, aggregated across all storage servers
|
|
class PerTagStatistics {
|
|
Optional<ThrottleApi::TagQuotaValue> quota;
|
|
Smoother transactionCounter;
|
|
Smoother perClientRate;
|
|
Smoother targetRate;
|
|
double transactionsLastAdded;
|
|
|
|
public:
|
|
explicit PerTagStatistics()
|
|
: transactionCounter(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_FOLDING_TIME),
|
|
perClientRate(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_FOLDING_TIME),
|
|
targetRate(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_FOLDING_TIME), transactionsLastAdded(now()) {}
|
|
|
|
Optional<ThrottleApi::TagQuotaValue> getQuota() const { return quota; }
|
|
|
|
void setQuota(ThrottleApi::TagQuotaValue quota) { this->quota = quota; }
|
|
|
|
void clearQuota() { quota = {}; }
|
|
|
|
void addTransactions(int count) {
|
|
transactionsLastAdded = now();
|
|
transactionCounter.addDelta(count);
|
|
}
|
|
|
|
double getTransactionRate() const { return transactionCounter.smoothRate(); }
|
|
|
|
ClientTagThrottleLimits updateAndGetPerClientLimit(double targetTps) {
|
|
auto newPerClientRate = std::max(
|
|
SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MIN_RATE,
|
|
std::min(targetTps, (targetTps / transactionCounter.smoothRate()) * perClientRate.smoothTotal()));
|
|
perClientRate.setTotal(newPerClientRate);
|
|
return ClientTagThrottleLimits(
|
|
std::max(perClientRate.smoothTotal(), SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MIN_RATE),
|
|
ClientTagThrottleLimits::NO_EXPIRATION);
|
|
}
|
|
|
|
double updateAndGetTargetLimit(double targetTps) {
|
|
targetRate.setTotal(targetTps);
|
|
return targetRate.smoothTotal();
|
|
}
|
|
|
|
bool recentTransactionsAdded() const {
|
|
return now() - transactionsLastAdded < SERVER_KNOBS->GLOBAL_TAG_THROTTLING_TAG_EXPIRE_AFTER;
|
|
}
|
|
};
|
|
|
|
struct StorageServerInfo {
|
|
Optional<Standalone<StringRef>> zoneId;
|
|
Optional<double> throttlingRatio;
|
|
};
|
|
|
|
Database db;
|
|
UID id;
|
|
int maxFallingBehind{ 0 };
|
|
uint64_t throttledTagChangeId{ 0 };
|
|
uint32_t lastBusyTagCount{ 0 };
|
|
|
|
std::unordered_map<UID, StorageServerInfo> ssInfos;
|
|
std::unordered_map<TransactionTag, PerTagStatistics> tagStatistics;
|
|
std::unordered_map<UID, std::unordered_map<TransactionTag, ThroughputCounters>> throughput;
|
|
|
|
// Returns the cost rate for the given tag on the given storage server
|
|
Optional<double> getCurrentCost(UID storageServerId, TransactionTag tag) const {
|
|
auto const tagToThroughputCounters = tryGet(throughput, storageServerId);
|
|
if (!tagToThroughputCounters.present()) {
|
|
return {};
|
|
}
|
|
auto const throughputCounter = tryGet(tagToThroughputCounters.get(), tag);
|
|
if (!throughputCounter.present()) {
|
|
return {};
|
|
}
|
|
return throughputCounter.get().getCost();
|
|
}
|
|
|
|
// Return the cost rate on the given storage server, summed across all tags
|
|
Optional<double> getCurrentCost(UID storageServerId) const {
|
|
auto tagToPerTagThroughput = tryGet(throughput, storageServerId);
|
|
if (!tagToPerTagThroughput.present()) {
|
|
return {};
|
|
}
|
|
double result = 0;
|
|
for (const auto& [tag, perTagThroughput] : tagToPerTagThroughput.get()) {
|
|
result += perTagThroughput.getCost();
|
|
}
|
|
return result;
|
|
}
|
|
|
|
// Return the cost rate for the given tag, summed across all storage servers
|
|
double getCurrentCost(TransactionTag tag) const {
|
|
double result{ 0.0 };
|
|
for (const auto& [id, _] : throughput) {
|
|
result += getCurrentCost(id, tag).orDefault(0);
|
|
}
|
|
// FIXME: Disabled due to noisy trace events. Fix the noise and reenabled
|
|
//TraceEvent("GlobalTagThrottler_GetCurrentCost").detail("Tag", printable(tag)).detail("Cost", result);
|
|
|
|
return result;
|
|
}
|
|
|
|
// For transactions with the provided tag, returns the average cost that gets associated with the provided storage
|
|
// server
|
|
Optional<double> getAverageTransactionCost(TransactionTag tag, UID storageServerId) const {
|
|
auto const cost = getCurrentCost(storageServerId, tag);
|
|
if (!cost.present()) {
|
|
return {};
|
|
}
|
|
auto const stats = tryGet(tagStatistics, tag);
|
|
if (!stats.present()) {
|
|
return {};
|
|
}
|
|
auto const transactionRate = stats.get().getTransactionRate();
|
|
// If there is less than GLOBAL_TAG_THROTTLING_MIN_TPS transactions per second, we do not have enough data
|
|
// to accurately compute an average transaction cost.
|
|
if (transactionRate < SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MIN_TPS) {
|
|
return {};
|
|
} else {
|
|
return std::max(static_cast<double>(CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE), cost.get() / transactionRate);
|
|
}
|
|
}
|
|
|
|
// For transactions with the provided tag, returns the average cost of all transactions
|
|
// accross the cluster. The minimum cost is one page. If the transaction rate is too low,
|
|
// return an empty Optional, because no accurate estimation can be made.
|
|
Optional<double> getAverageTransactionCost(TransactionTag tag, TraceEvent& te) const {
|
|
auto const cost = getCurrentCost(tag);
|
|
auto const stats = tryGet(tagStatistics, tag);
|
|
if (!stats.present()) {
|
|
return CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
}
|
|
auto const transactionRate = stats.get().getTransactionRate();
|
|
te.detail("TransactionRate", transactionRate);
|
|
te.detail("Cost", cost);
|
|
if (transactionRate < SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MIN_TPS) {
|
|
return {};
|
|
} else {
|
|
return std::max(static_cast<double>(CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE), cost / transactionRate);
|
|
}
|
|
}
|
|
|
|
// Returns the list of all tags performing meaningful work on the given storage server
|
|
std::vector<TransactionTag> getTagsAffectingStorageServer(UID storageServerId) const {
|
|
std::vector<TransactionTag> result;
|
|
auto const tagToThroughputCounters = tryGet(throughput, storageServerId);
|
|
if (!tagToThroughputCounters.present()) {
|
|
return {};
|
|
} else {
|
|
result.reserve(tagToThroughputCounters.get().size());
|
|
for (const auto& [t, _] : tagToThroughputCounters.get()) {
|
|
result.push_back(t);
|
|
}
|
|
}
|
|
return result;
|
|
}
|
|
|
|
Optional<double> getQuota(TransactionTag tag, LimitType limitType) const {
|
|
auto const stats = tryGet(tagStatistics, tag);
|
|
if (!stats.present()) {
|
|
return {};
|
|
}
|
|
auto const quota = stats.get().getQuota();
|
|
if (!quota.present()) {
|
|
return {};
|
|
}
|
|
if (limitType == LimitType::TOTAL) {
|
|
return quota.get().totalQuota;
|
|
} else {
|
|
return quota.get().reservedQuota;
|
|
}
|
|
}
|
|
|
|
// Of all tags meaningfully performing workload on the given storage server,
|
|
// returns the ratio of total quota allocated to the specified tag
|
|
double getQuotaRatio(TransactionTagRef tag, UID storageServerId) const {
|
|
double sumQuota{ 0.0 };
|
|
double tagQuota{ 0.0 };
|
|
auto const tagsAffectingStorageServer = getTagsAffectingStorageServer(storageServerId);
|
|
for (const auto& t : tagsAffectingStorageServer) {
|
|
auto const tQuota = getQuota(t, LimitType::TOTAL);
|
|
sumQuota += tQuota.orDefault(0);
|
|
if (t.compare(tag) == 0) {
|
|
tagQuota = tQuota.orDefault(0);
|
|
}
|
|
}
|
|
if (tagQuota == 0.0) {
|
|
return 0;
|
|
}
|
|
ASSERT_GT(sumQuota, 0.0);
|
|
return tagQuota / sumQuota;
|
|
}
|
|
|
|
// Returns the desired cost for a storage server, based on its current
|
|
// cost and throttling ratio
|
|
Optional<double> getLimitingCost(UID storageServerId) const {
|
|
auto const ssInfo = tryGet(ssInfos, storageServerId);
|
|
Optional<double> const throttlingRatio = ssInfo.present() ? ssInfo.get().throttlingRatio : Optional<double>{};
|
|
Optional<double> const currentCost = getCurrentCost(storageServerId);
|
|
if (!throttlingRatio.present() || !currentCost.present()) {
|
|
return {};
|
|
}
|
|
return throttlingRatio.get() * currentCost.get();
|
|
}
|
|
|
|
// For a given storage server and tag combination, return the limiting transaction rate.
|
|
Optional<double> getLimitingTps(UID storageServerId, TransactionTag tag) const {
|
|
auto const quotaRatio = getQuotaRatio(tag, storageServerId);
|
|
Optional<double> const limitingCost = getLimitingCost(storageServerId);
|
|
Optional<double> const averageTransactionCost = getAverageTransactionCost(tag, storageServerId);
|
|
if (!limitingCost.present() || !averageTransactionCost.present()) {
|
|
return {};
|
|
}
|
|
|
|
auto const limitingCostForTag = limitingCost.get() * quotaRatio;
|
|
return limitingCostForTag / averageTransactionCost.get();
|
|
}
|
|
|
|
// Return the limiting transaction rate, aggregated across all storage servers.
|
|
// The limits from the worst maxFallingBehind zones are
|
|
// ignored, because we do not non-workload related issues (e.g. slow disks)
|
|
// to affect tag throttling. If more than maxFallingBehind zones are at
|
|
// or near saturation, this indicates that throttling should take place.
|
|
Optional<double> getLimitingTps(TransactionTag tag) const {
|
|
// TODO: The algorithm for ignoring the worst zones can be made more efficient
|
|
std::unordered_map<Optional<Standalone<StringRef>>, double> zoneIdToLimitingTps;
|
|
for (const auto& [id, ssInfo] : ssInfos) {
|
|
auto const limitingTpsForSS = getLimitingTps(id, tag);
|
|
if (limitingTpsForSS.present()) {
|
|
auto it = zoneIdToLimitingTps.find(ssInfo.zoneId);
|
|
if (it != zoneIdToLimitingTps.end()) {
|
|
auto& limitingTpsForZone = it->second;
|
|
limitingTpsForZone = std::min<double>(limitingTpsForZone, limitingTpsForSS.get());
|
|
} else {
|
|
zoneIdToLimitingTps[ssInfo.zoneId] = limitingTpsForSS.get();
|
|
}
|
|
}
|
|
}
|
|
if (zoneIdToLimitingTps.size() <= maxFallingBehind) {
|
|
return {};
|
|
} else {
|
|
std::vector<double> zoneLimits;
|
|
for (const auto& [_, limit] : zoneIdToLimitingTps) {
|
|
zoneLimits.push_back(limit);
|
|
}
|
|
std::nth_element(zoneLimits.begin(), zoneLimits.begin() + maxFallingBehind, zoneLimits.end());
|
|
return zoneLimits[maxFallingBehind];
|
|
}
|
|
}
|
|
|
|
void removeUnseenQuotas(std::unordered_set<TransactionTag> const& tagsWithQuota) {
|
|
for (auto& [tag, stats] : tagStatistics) {
|
|
if (!tagsWithQuota.count(tag)) {
|
|
stats.clearQuota();
|
|
}
|
|
}
|
|
}
|
|
|
|
ACTOR static Future<Void> monitorThrottlingChanges(GlobalTagThrottlerImpl* self) {
|
|
state std::unordered_set<TransactionTag> tagsWithQuota;
|
|
|
|
loop {
|
|
state ReadYourWritesTransaction tr(self->db);
|
|
loop {
|
|
try {
|
|
tr.setOption(FDBTransactionOptions::ACCESS_SYSTEM_KEYS);
|
|
tr.setOption(FDBTransactionOptions::PRIORITY_SYSTEM_IMMEDIATE);
|
|
|
|
tagsWithQuota.clear();
|
|
state RangeResult currentQuotas = wait(tr.getRange(tagQuotaKeys, CLIENT_KNOBS->TOO_MANY));
|
|
TraceEvent("GlobalTagThrottler_ReadCurrentQuotas", self->id).detail("Size", currentQuotas.size());
|
|
for (auto const kv : currentQuotas) {
|
|
auto const tag = kv.key.removePrefix(tagQuotaPrefix);
|
|
auto const quota = ThrottleApi::TagQuotaValue::fromValue(kv.value);
|
|
self->tagStatistics[tag].setQuota(quota);
|
|
tagsWithQuota.insert(tag);
|
|
}
|
|
self->removeUnseenQuotas(tagsWithQuota);
|
|
self->removeExpiredTags();
|
|
++self->throttledTagChangeId;
|
|
wait(delay(5.0));
|
|
break;
|
|
} catch (Error& e) {
|
|
TraceEvent("GlobalTagThrottler_MonitoringChangesError", self->id).error(e);
|
|
wait(tr.onError(e));
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
Optional<double> getTargetTps(TransactionTag tag, bool& isBusy, TraceEvent& te) {
|
|
auto const limitingTps = getLimitingTps(tag);
|
|
auto const averageTransactionCost = getAverageTransactionCost(tag, te);
|
|
auto const totalQuota = getQuota(tag, LimitType::TOTAL);
|
|
auto const reservedQuota = getQuota(tag, LimitType::RESERVED);
|
|
if (!averageTransactionCost.present() || !totalQuota.present() || !reservedQuota.present()) {
|
|
return {};
|
|
}
|
|
auto const desiredTps = totalQuota.get() / averageTransactionCost.get();
|
|
auto const reservedTps = reservedQuota.get() / averageTransactionCost.get();
|
|
auto const targetTps = getMax(reservedTps, getMin(desiredTps, limitingTps));
|
|
|
|
isBusy = limitingTps.present() && limitingTps.get() < desiredTps;
|
|
|
|
te.detail("Tag", printable(tag))
|
|
.detail("TargetTps", targetTps)
|
|
.detail("AverageTransactionCost", averageTransactionCost)
|
|
.detail("LimitingTps", limitingTps)
|
|
.detail("ReservedTps", reservedTps)
|
|
.detail("DesiredTps", desiredTps)
|
|
.detail("NumStorageServers", throughput.size())
|
|
.detail("TotalQuota", totalQuota)
|
|
.detail("ReservedQuota", reservedQuota);
|
|
|
|
return targetTps;
|
|
}
|
|
|
|
public:
|
|
GlobalTagThrottlerImpl(Database db, UID id, int maxFallingBehind)
|
|
: db(db), id(id), maxFallingBehind(maxFallingBehind) {}
|
|
Future<Void> monitorThrottlingChanges() { return monitorThrottlingChanges(this); }
|
|
void addRequests(TransactionTag tag, int count) {
|
|
auto it = tagStatistics.find(tag);
|
|
if (it == tagStatistics.end()) {
|
|
if (tagStatistics.size() == SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MAX_TAGS_TRACKED) {
|
|
CODE_PROBE(true,
|
|
"Global tag throttler ignoring transactions because maximum number of trackable tags has "
|
|
"been reached");
|
|
TraceEvent("GlobalTagThrottler_IgnoringRequests")
|
|
.suppressFor(60.0)
|
|
.detail("Tag", printable(tag))
|
|
.detail("Count", count);
|
|
} else {
|
|
tagStatistics[tag].addTransactions(static_cast<double>(count));
|
|
}
|
|
} else {
|
|
it->second.addTransactions(static_cast<double>(count));
|
|
}
|
|
}
|
|
uint64_t getThrottledTagChangeId() const { return throttledTagChangeId; }
|
|
|
|
TransactionTagMap<double> getProxyRates(int numProxies) {
|
|
TransactionTagMap<double> result;
|
|
lastBusyTagCount = 0;
|
|
|
|
for (auto& [tag, stats] : tagStatistics) {
|
|
// Currently there is no differentiation between batch priority and default priority transactions
|
|
TraceEvent te("GlobalTagThrottler_GotRate", id);
|
|
bool isBusy{ false };
|
|
auto const targetTps = getTargetTps(tag, isBusy, te);
|
|
if (isBusy) {
|
|
++lastBusyTagCount;
|
|
}
|
|
if (targetTps.present()) {
|
|
auto const smoothedTargetTps = stats.updateAndGetTargetLimit(targetTps.get());
|
|
te.detail("SmoothedTargetTps", smoothedTargetTps).detail("NumProxies", numProxies);
|
|
result[tag] = std::max(1.0, smoothedTargetTps / numProxies);
|
|
} else {
|
|
te.disable();
|
|
}
|
|
}
|
|
|
|
return result;
|
|
}
|
|
|
|
PrioritizedTransactionTagMap<ClientTagThrottleLimits> getClientRates() {
|
|
PrioritizedTransactionTagMap<ClientTagThrottleLimits> result;
|
|
lastBusyTagCount = 0;
|
|
|
|
for (auto& [tag, stats] : tagStatistics) {
|
|
// Currently there is no differentiation between batch priority and default priority transactions
|
|
bool isBusy{ false };
|
|
TraceEvent te("GlobalTagThrottler_GotClientRate", id);
|
|
auto const targetTps = getTargetTps(tag, isBusy, te);
|
|
|
|
if (isBusy) {
|
|
++lastBusyTagCount;
|
|
}
|
|
|
|
if (targetTps.present()) {
|
|
auto const clientRate = stats.updateAndGetPerClientLimit(targetTps.get());
|
|
result[TransactionPriority::BATCH][tag] = result[TransactionPriority::DEFAULT][tag] = clientRate;
|
|
te.detail("ClientTps", clientRate.tpsRate);
|
|
} else {
|
|
te.disable();
|
|
}
|
|
}
|
|
return result;
|
|
}
|
|
|
|
int64_t autoThrottleCount() const {
|
|
int64_t result{ 0 };
|
|
for (const auto& [tag, stats] : tagStatistics) {
|
|
if (stats.getQuota().present()) {
|
|
++result;
|
|
}
|
|
}
|
|
return result;
|
|
}
|
|
uint32_t busyReadTagCount() const { return lastBusyTagCount; }
|
|
uint32_t busyWriteTagCount() const { return lastBusyTagCount; }
|
|
int64_t manualThrottleCount() const { return 0; }
|
|
|
|
Future<Void> tryUpdateAutoThrottling(StorageQueueInfo const& ss) {
|
|
auto& ssInfo = ssInfos[ss.id];
|
|
ssInfo.throttlingRatio = ss.getTagThrottlingRatio(SERVER_KNOBS->TARGET_BYTES_PER_STORAGE_SERVER,
|
|
SERVER_KNOBS->SPRING_BYTES_STORAGE_SERVER);
|
|
ssInfo.zoneId = ss.locality.zoneId();
|
|
|
|
for (const auto& busyReadTag : ss.busiestReadTags) {
|
|
if (tagStatistics.find(busyReadTag.tag) != tagStatistics.end()) {
|
|
throughput[ss.id][busyReadTag.tag].updateCost(busyReadTag.rate, OpType::READ);
|
|
}
|
|
}
|
|
for (const auto& busyWriteTag : ss.busiestWriteTags) {
|
|
if (tagStatistics.find(busyWriteTag.tag) != tagStatistics.end()) {
|
|
throughput[ss.id][busyWriteTag.tag].updateCost(busyWriteTag.rate, OpType::WRITE);
|
|
}
|
|
}
|
|
return Void();
|
|
}
|
|
|
|
void setQuota(TransactionTagRef tag, ThrottleApi::TagQuotaValue const& tagQuotaValue) {
|
|
tagStatistics[tag].setQuota(tagQuotaValue);
|
|
}
|
|
|
|
void removeQuota(TransactionTagRef tag) { tagStatistics[tag].clearQuota(); }
|
|
|
|
void removeExpiredTags() {
|
|
for (auto it = tagStatistics.begin(); it != tagStatistics.end();) {
|
|
const auto& [tag, stats] = *it;
|
|
if (!stats.recentTransactionsAdded()) {
|
|
for (auto& [ss, tagToCounters] : throughput) {
|
|
tagToCounters.erase(tag);
|
|
}
|
|
it = tagStatistics.erase(it);
|
|
} else {
|
|
++it;
|
|
}
|
|
}
|
|
}
|
|
|
|
uint32_t tagsTracked() const { return tagStatistics.size(); }
|
|
};
|
|
|
|
GlobalTagThrottler::GlobalTagThrottler(Database db, UID id, int maxFallingBehind)
|
|
: impl(PImpl<GlobalTagThrottlerImpl>::create(db, id, maxFallingBehind)) {}
|
|
|
|
GlobalTagThrottler::~GlobalTagThrottler() = default;
|
|
|
|
Future<Void> GlobalTagThrottler::monitorThrottlingChanges() {
|
|
return impl->monitorThrottlingChanges();
|
|
}
|
|
void GlobalTagThrottler::addRequests(TransactionTag tag, int count) {
|
|
return impl->addRequests(tag, count);
|
|
}
|
|
uint64_t GlobalTagThrottler::getThrottledTagChangeId() const {
|
|
return impl->getThrottledTagChangeId();
|
|
}
|
|
PrioritizedTransactionTagMap<ClientTagThrottleLimits> GlobalTagThrottler::getClientRates() {
|
|
return impl->getClientRates();
|
|
}
|
|
TransactionTagMap<double> GlobalTagThrottler::getProxyRates(int numProxies) {
|
|
return impl->getProxyRates(numProxies);
|
|
}
|
|
int64_t GlobalTagThrottler::autoThrottleCount() const {
|
|
return impl->autoThrottleCount();
|
|
}
|
|
uint32_t GlobalTagThrottler::busyReadTagCount() const {
|
|
return impl->busyReadTagCount();
|
|
}
|
|
uint32_t GlobalTagThrottler::busyWriteTagCount() const {
|
|
return impl->busyWriteTagCount();
|
|
}
|
|
int64_t GlobalTagThrottler::manualThrottleCount() const {
|
|
return impl->manualThrottleCount();
|
|
}
|
|
bool GlobalTagThrottler::isAutoThrottlingEnabled() const {
|
|
return true;
|
|
}
|
|
Future<Void> GlobalTagThrottler::tryUpdateAutoThrottling(StorageQueueInfo const& ss) {
|
|
return impl->tryUpdateAutoThrottling(ss);
|
|
}
|
|
|
|
void GlobalTagThrottler::setQuota(TransactionTagRef tag, ThrottleApi::TagQuotaValue const& tagQuotaValue) {
|
|
return impl->setQuota(tag, tagQuotaValue);
|
|
}
|
|
|
|
void GlobalTagThrottler::removeQuota(TransactionTagRef tag) {
|
|
return impl->removeQuota(tag);
|
|
}
|
|
|
|
uint32_t GlobalTagThrottler::tagsTracked() const {
|
|
return impl->tagsTracked();
|
|
}
|
|
|
|
void GlobalTagThrottler::removeExpiredTags() {
|
|
return impl->removeExpiredTags();
|
|
}
|
|
|
|
namespace {
|
|
|
|
enum class LimitType { RESERVED, TOTAL };
|
|
enum class OpType { READ, WRITE };
|
|
|
|
Optional<double> getTPSLimit(GlobalTagThrottler& globalTagThrottler, TransactionTag tag) {
|
|
auto clientRates = globalTagThrottler.getClientRates();
|
|
auto it1 = clientRates.find(TransactionPriority::DEFAULT);
|
|
if (it1 != clientRates.end()) {
|
|
auto it2 = it1->second.find(tag);
|
|
if (it2 != it1->second.end()) {
|
|
return it2->second.tpsRate;
|
|
}
|
|
}
|
|
return {};
|
|
}
|
|
|
|
class MockStorageServer {
|
|
class Cost {
|
|
Smoother smoother;
|
|
|
|
public:
|
|
Cost() : smoother(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_FOLDING_TIME) {}
|
|
Cost& operator+=(double delta) {
|
|
smoother.addDelta(delta);
|
|
return *this;
|
|
}
|
|
double smoothRate() const { return smoother.smoothRate(); }
|
|
};
|
|
|
|
UID id;
|
|
// pages/second that this storage server can handle
|
|
double capacity;
|
|
std::map<TransactionTag, Cost> readCosts, writeCosts;
|
|
Cost totalReadCost, totalWriteCost;
|
|
|
|
public:
|
|
explicit MockStorageServer(UID id, double capacity) : id(id), capacity(capacity) { ASSERT_GT(capacity, 0); }
|
|
void addReadCost(TransactionTag tag, double cost) {
|
|
readCosts[tag] += cost;
|
|
totalReadCost += cost;
|
|
}
|
|
void addWriteCost(TransactionTag tag, double cost) {
|
|
writeCosts[tag] += cost;
|
|
totalWriteCost += cost;
|
|
}
|
|
|
|
void setCapacity(double value) { capacity = value; }
|
|
|
|
StorageQueueInfo getStorageQueueInfo() const {
|
|
StorageQueueInfo result(id, LocalityData({}, Value(id.toString()), {}, {}));
|
|
for (const auto& [tag, readCost] : readCosts) {
|
|
double fractionalBusyness{ 0.0 }; // unused for global tag throttling
|
|
result.busiestReadTags.emplace_back(tag, readCost.smoothRate(), fractionalBusyness);
|
|
}
|
|
for (const auto& [tag, writeCost] : writeCosts) {
|
|
double fractionalBusyness{ 0.0 }; // unused for global tag throttling
|
|
result.busiestWriteTags.emplace_back(tag, writeCost.smoothRate(), fractionalBusyness);
|
|
}
|
|
result.lastReply.bytesInput = ((totalReadCost.smoothRate() + totalWriteCost.smoothRate()) /
|
|
(capacity * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE)) *
|
|
SERVER_KNOBS->TARGET_BYTES_PER_STORAGE_SERVER;
|
|
return result;
|
|
}
|
|
};
|
|
|
|
class StorageServerCollection {
|
|
std::vector<MockStorageServer> storageServers;
|
|
|
|
public:
|
|
StorageServerCollection(size_t size, double capacity) {
|
|
ASSERT_GT(size, 0);
|
|
storageServers.reserve(size);
|
|
for (int i = 0; i < size; ++i) {
|
|
storageServers.emplace_back(UID(i, i), capacity);
|
|
}
|
|
}
|
|
|
|
void addCost(TransactionTag tag,
|
|
double pagesPerSecond,
|
|
std::vector<int> const& storageServerIndices,
|
|
OpType opType) {
|
|
if (storageServerIndices.empty()) {
|
|
auto costPerSS = CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE * (pagesPerSecond / storageServers.size());
|
|
if (opType == OpType::WRITE) {
|
|
costPerSS *= CLIENT_KNOBS->GLOBAL_TAG_THROTTLING_RW_FUNGIBILITY_RATIO;
|
|
}
|
|
for (auto& storageServer : storageServers) {
|
|
if (opType == OpType::READ) {
|
|
storageServer.addReadCost(tag, costPerSS);
|
|
} else {
|
|
storageServer.addWriteCost(tag, costPerSS);
|
|
}
|
|
}
|
|
} else {
|
|
auto const costPerSS =
|
|
CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE * (pagesPerSecond / storageServerIndices.size());
|
|
for (auto i : storageServerIndices) {
|
|
if (opType == OpType::READ) {
|
|
storageServers[i].addReadCost(tag, costPerSS);
|
|
} else {
|
|
storageServers[i].addWriteCost(tag, costPerSS);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
void setCapacity(int index, double value) { storageServers[index].setCapacity(value); }
|
|
|
|
std::vector<StorageQueueInfo> getStorageQueueInfos() const {
|
|
std::vector<StorageQueueInfo> result;
|
|
result.reserve(storageServers.size());
|
|
for (const auto& storageServer : storageServers) {
|
|
result.push_back(storageServer.getStorageQueueInfo());
|
|
}
|
|
return result;
|
|
}
|
|
};
|
|
|
|
ACTOR Future<Void> runClient(GlobalTagThrottler* globalTagThrottler,
|
|
StorageServerCollection* storageServers,
|
|
TransactionTag tag,
|
|
double tpsRate,
|
|
double costPerTransaction,
|
|
OpType opType,
|
|
std::vector<int> storageServerIndices = std::vector<int>()) {
|
|
loop {
|
|
auto tpsLimit = getTPSLimit(*globalTagThrottler, tag);
|
|
state double enforcedRate = tpsLimit.present() ? std::min<double>(tpsRate, tpsLimit.get()) : tpsRate;
|
|
wait(delay(1 / enforcedRate));
|
|
storageServers->addCost(tag, costPerTransaction, storageServerIndices, opType);
|
|
globalTagThrottler->addRequests(tag, 1);
|
|
}
|
|
}
|
|
|
|
ACTOR template <class Check>
|
|
Future<Void> monitorActor(GlobalTagThrottler* globalTagThrottler, Check check) {
|
|
state int successes = 0;
|
|
loop {
|
|
wait(delay(1.0));
|
|
if (check(*globalTagThrottler)) {
|
|
// Wait for 10 consecutive successes so we're certain
|
|
// than a stable equilibrium has been reached
|
|
if (++successes == 10) {
|
|
return Void();
|
|
}
|
|
} else {
|
|
successes = 0;
|
|
}
|
|
}
|
|
}
|
|
|
|
bool isNear(double a, double b) {
|
|
return abs(a - b) < 0.1 * a;
|
|
}
|
|
|
|
bool isNear(Optional<double> a, Optional<double> b) {
|
|
if (a.present()) {
|
|
return b.present() && isNear(a.get(), b.get());
|
|
} else {
|
|
return !b.present();
|
|
}
|
|
}
|
|
|
|
bool targetRateIsNear(GlobalTagThrottler& globalTagThrottler, TransactionTag tag, Optional<double> expected) {
|
|
Optional<double> rate;
|
|
auto targetRates = globalTagThrottler.getProxyRates(1);
|
|
auto it = targetRates.find(tag);
|
|
if (it != targetRates.end()) {
|
|
rate = it->second;
|
|
}
|
|
TraceEvent("GlobalTagThrottling_RateMonitor")
|
|
.detail("Tag", tag)
|
|
.detail("CurrentTPSRate", rate)
|
|
.detail("ExpectedTPSRate", expected);
|
|
return isNear(rate, expected);
|
|
}
|
|
|
|
bool clientRateIsNear(GlobalTagThrottler& globalTagThrottler, TransactionTag tag, Optional<double> expected) {
|
|
Optional<double> rate;
|
|
auto clientRates = globalTagThrottler.getClientRates();
|
|
auto it1 = clientRates.find(TransactionPriority::DEFAULT);
|
|
if (it1 != clientRates.end()) {
|
|
auto it2 = it1->second.find(tag);
|
|
if (it2 != it1->second.end()) {
|
|
rate = it2->second.tpsRate;
|
|
}
|
|
}
|
|
TraceEvent("GlobalTagThrottling_ClientRateMonitor")
|
|
.detail("Tag", tag)
|
|
.detail("CurrentTPSRate", rate)
|
|
.detail("ExpectedTPSRate", expected);
|
|
return isNear(rate, expected);
|
|
}
|
|
|
|
ACTOR Future<Void> updateGlobalTagThrottler(GlobalTagThrottler* globalTagThrottler,
|
|
StorageServerCollection const* storageServers) {
|
|
loop {
|
|
wait(delay(1.0));
|
|
auto const storageQueueInfos = storageServers->getStorageQueueInfos();
|
|
for (const auto& sq : storageQueueInfos) {
|
|
globalTagThrottler->tryUpdateAutoThrottling(sq);
|
|
}
|
|
}
|
|
}
|
|
|
|
} // namespace
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second.
|
|
// Client attempts 5 6-byte read transactions per second.
|
|
// Limit should adjust to allow 100/6 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/Simple") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor =
|
|
monitorActor(&globalTagThrottler, [testTag](auto& gtt) { return targetRateIsNear(gtt, testTag, 100.0 / 6.0); });
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second.
|
|
// Client attempts 5 6-byte write transactions per second.
|
|
// Limit should adjust to allow 100/(6*<fungibility_ratio>) transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/WriteThrottling") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::WRITE);
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag, 100.0 / (6.0 * CLIENT_KNOBS->GLOBAL_TAG_THROTTLING_RW_FUNGIBILITY_RATIO));
|
|
});
|
|
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second for each tag.
|
|
// 2 clients each attempt 5 6-byte read transactions per second.
|
|
// Both limits should adjust to allow 100/6 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/MultiTagThrottling") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag1 = "sampleTag1"_sr;
|
|
TransactionTag testTag2 = "sampleTag2"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag1, tagQuotaValue);
|
|
globalTagThrottler.setQuota(testTag2, tagQuotaValue);
|
|
state std::vector<Future<Void>> futures;
|
|
state std::vector<Future<Void>> monitorFutures;
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag1, 5.0, 6.0, OpType::READ));
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag2, 5.0, 6.0, OpType::READ));
|
|
futures.push_back(updateGlobalTagThrottler(&globalTagThrottler, &storageServers));
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag1, testTag2](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag1, 100.0 / 6.0) && targetRateIsNear(gtt, testTag2, 100.0 / 6.0);
|
|
});
|
|
wait(timeoutError(waitForAny(futures) || monitor, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second.
|
|
// Client attempts 20 10-byte read transactions per second.
|
|
// Limit should adjust to allow 100/10 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/AttemptWorkloadAboveQuota") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 20.0, 10.0, OpType::READ);
|
|
state Future<Void> monitor =
|
|
monitorActor(&globalTagThrottler, [testTag](auto& gtt) { return targetRateIsNear(gtt, testTag, 10.0); });
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second.
|
|
// 2 clients each attempt 5 6-byte transactions per second.
|
|
// Limit should adjust to allow 100/6 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/MultiClientThrottling") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::READ);
|
|
state Future<Void> client2 = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag, 100.0 / 6.0) && clientRateIsNear(gtt, testTag, 100.0 / 6.0);
|
|
});
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || client2 || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second.
|
|
// 2 clients each attempt 20 10-page transactions per second.
|
|
// Target rate should adjust to allow 100/10 transactions per second.
|
|
// Each client is throttled to only perform (100/10)/2 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/MultiClientThrottling2") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 20.0, 10.0, OpType::READ);
|
|
state Future<Void> client2 = runClient(&globalTagThrottler, &storageServers, testTag, 20.0, 10.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag, 10.0) && clientRateIsNear(gtt, testTag, 5.0);
|
|
});
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota set to 100 pages/second.
|
|
// One client attempts 5 5-page read transactions per second.
|
|
// Another client attempts 25 5-page read transactions per second.
|
|
// Target rate should adjust to allow 100/5 transactions per second.
|
|
// This 20 transactions/second limit is split with a distribution of (5, 15) between the 2 clients.
|
|
TEST_CASE("/GlobalTagThrottler/SkewedMultiClientThrottling") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 5.0, OpType::READ);
|
|
state Future<Void> client2 = runClient(&globalTagThrottler, &storageServers, testTag, 25.0, 5.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag, 20.0) && clientRateIsNear(gtt, testTag, 15.0);
|
|
});
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota is initially set to 100 pages/second.
|
|
// Client attempts 5 6-page transactions per second.
|
|
// Test that the tag throttler can reach equilibrium, then adjust to a new equilibrium once the quota is changed
|
|
// Target rate should adjust to allow 100/6 transactions per second.
|
|
// Total quota is modified to 50 pages/second.
|
|
// Target rate should adjust to allow 50/6 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/UpdateQuota") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
state TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(
|
|
&globalTagThrottler, [](auto& gtt) { return targetRateIsNear(gtt, "sampleTag1"_sr, 100.0 / 6.0); });
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
tagQuotaValue.totalQuota = 50 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
monitor =
|
|
monitorActor(&globalTagThrottler, [](auto& gtt) { return targetRateIsNear(gtt, "sampleTag1"_sr, 50.0 / 6.0); });
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 100 pages/second each.
|
|
// Total quota is initially set to 100 pages/second.
|
|
// Client attempts 5 6-page read transactions per second.
|
|
// Target limit adjusts to allow 100/6 transactions per second.
|
|
// Then Quota is removed.
|
|
// Target limit is removed as a result.
|
|
TEST_CASE("/GlobalTagThrottler/RemoveQuota") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
state TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(
|
|
&globalTagThrottler, [](auto& gtt) { return targetRateIsNear(gtt, "sampleTag1"_sr, 100.0 / 6.0); });
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
globalTagThrottler.removeQuota(testTag);
|
|
monitor = monitorActor(&globalTagThrottler, [](auto& gtt) { return targetRateIsNear(gtt, "sampleTag1"_sr, {}); });
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 5 pages/second each.
|
|
// Total quota is set to 100 pages/second.
|
|
// Client attempts 10 6-page transactions per second
|
|
// Target is adjusted to 50/6 transactions per second, to match the total capacity all storage servers.
|
|
TEST_CASE("/GlobalTagThrottler/ActiveThrottling") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 5);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 10.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag, 50 / 6.0) && gtt.busyReadTagCount() == 1;
|
|
});
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can handle 5 pages/second each.
|
|
// Total quota is set to 50 pages/second for one tag, 100 pages/second for another.
|
|
// For each tag, a client attempts to execute 10 6-page read transactions per second.
|
|
// Target rates are adjusted to utilize the full 50 pages/second capacity of the
|
|
// add storage servers. The two tags receive this capacity with a 2:1 ratio,
|
|
// matching the ratio of their total quotas.
|
|
TEST_CASE("/GlobalTagThrottler/MultiTagActiveThrottling") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 5);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue1;
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue2;
|
|
TransactionTag testTag1 = "sampleTag1"_sr;
|
|
TransactionTag testTag2 = "sampleTag2"_sr;
|
|
tagQuotaValue1.totalQuota = 50 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
tagQuotaValue2.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag1, tagQuotaValue1);
|
|
globalTagThrottler.setQuota(testTag2, tagQuotaValue2);
|
|
std::vector<Future<Void>> futures;
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag1, 10.0, 6.0, OpType::READ));
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag2, 10.0, 6.0, OpType::READ));
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag1, testTag2](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag1, (50 / 6.0) / 3) && targetRateIsNear(gtt, testTag2, 2 * (50 / 6.0) / 3) &&
|
|
gtt.busyReadTagCount() == 2;
|
|
});
|
|
futures.push_back(updateGlobalTagThrottler(&globalTagThrottler, &storageServers));
|
|
wait(timeoutError(waitForAny(futures) || monitor, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 3 storage servers can handle 50 pages/second each.
|
|
// Total quota is set to 100 pages/second for each tag.
|
|
// Each client attempts 10 6-page read transactions per second.
|
|
// This workload is sent to 2 storage servers per client (with an overlap of one storage server).
|
|
// Target rates for both tags are adjusted to 50/6 transactions per second to match the throughput
|
|
// that the busiest server can handle.
|
|
TEST_CASE("/GlobalTagThrottler/MultiTagActiveThrottling2") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(3, 50);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue1;
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue2;
|
|
TransactionTag testTag1 = "sampleTag1"_sr;
|
|
TransactionTag testTag2 = "sampleTag2"_sr;
|
|
tagQuotaValue1.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
tagQuotaValue2.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag1, tagQuotaValue1);
|
|
globalTagThrottler.setQuota(testTag2, tagQuotaValue2);
|
|
std::vector<Future<Void>> futures;
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag1, 10.0, 6.0, OpType::READ, { 0, 1 }));
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag2, 10.0, 6.0, OpType::READ, { 1, 2 }));
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag1, testTag2](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag1, 50 / 6.0) && targetRateIsNear(gtt, testTag2, 50 / 6.0) &&
|
|
gtt.busyReadTagCount() == 2;
|
|
});
|
|
futures.push_back(updateGlobalTagThrottler(&globalTagThrottler, &storageServers));
|
|
wait(timeoutError(waitForAny(futures) || monitor, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 3 storage servers can handle 50 pages/second each.
|
|
// Total quota is set to 100 pages/second for each tag.
|
|
// One client attempts 10 6-page read transactions per second, all directed towards a single storage server.
|
|
// Another client, using a different tag, attempts 10 6-page read transactions split across the other two storage
|
|
// servers. Target rates adjust to 50/6 and 100/6 transactions per second for the two clients, based on the capacities
|
|
// of the
|
|
// storage servers being accessed.
|
|
TEST_CASE("/GlobalTagThrottler/MultiTagActiveThrottling3") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(3, 50);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue1;
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue2;
|
|
TransactionTag testTag1 = "sampleTag1"_sr;
|
|
TransactionTag testTag2 = "sampleTag2"_sr;
|
|
tagQuotaValue1.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
tagQuotaValue2.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag1, tagQuotaValue1);
|
|
globalTagThrottler.setQuota(testTag2, tagQuotaValue2);
|
|
std::vector<Future<Void>> futures;
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag1, 10.0, 6.0, OpType::READ, { 0 }));
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, testTag2, 10.0, 6.0, OpType::READ, { 1, 2 }));
|
|
state Future<Void> monitor = monitorActor(&globalTagThrottler, [testTag1, testTag2](auto& gtt) {
|
|
return targetRateIsNear(gtt, testTag1, 50 / 6.0) && targetRateIsNear(gtt, testTag2, 100 / 6.0) &&
|
|
gtt.busyReadTagCount() == 1;
|
|
});
|
|
futures.push_back(updateGlobalTagThrottler(&globalTagThrottler, &storageServers));
|
|
wait(timeoutError(waitForAny(futures) || monitor, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// 10 storage servers can serve 5 pages/second each.
|
|
// Total quota is set to 100 pages/second.
|
|
// Reserved quota is set to 70 pages/second.
|
|
// A client attempts to execute 10 6-page read transactions per second.
|
|
// Despite the storage server only having capacity to serve 50/6 transactions per second,
|
|
// the reserved quota will ensure the target rate adjusts to 70/6 transactions per second.
|
|
TEST_CASE("/GlobalTagThrottler/ReservedQuota") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 5);
|
|
state ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
tagQuotaValue.reservedQuota = 70 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 10.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor =
|
|
monitorActor(&globalTagThrottler, [testTag](auto& gtt) { return targetRateIsNear(gtt, testTag, 70 / 6.0); });
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|
|
|
|
// Test that tags are expired iff a sufficient amount of time has passed since the
|
|
// last transaction with that tag
|
|
TEST_CASE("/GlobalTagThrottler/ExpireTags") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 5);
|
|
TransactionTag testTag = "sampleTag1"_sr;
|
|
|
|
state Future<Void> client =
|
|
timeout(runClient(&globalTagThrottler, &storageServers, testTag, 10.0, 6.0, OpType::READ), 60.0, Void());
|
|
state Future<Void> updater = timeout(updateGlobalTagThrottler(&globalTagThrottler, &storageServers), 60.0, Void());
|
|
wait(client && updater);
|
|
client.cancel();
|
|
updater.cancel();
|
|
ASSERT_EQ(globalTagThrottler.tagsTracked(), 1);
|
|
globalTagThrottler.removeExpiredTags();
|
|
ASSERT_EQ(globalTagThrottler.tagsTracked(), 1);
|
|
wait(delay(SERVER_KNOBS->GLOBAL_TAG_THROTTLING_TAG_EXPIRE_AFTER + 1.0));
|
|
ASSERT_EQ(globalTagThrottler.tagsTracked(), 1);
|
|
globalTagThrottler.removeExpiredTags();
|
|
ASSERT_EQ(globalTagThrottler.tagsTracked(), 0);
|
|
return Void();
|
|
}
|
|
|
|
// Test that the number of tags tracked does not grow beyond SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MAX_TAGS_TRACKED
|
|
TEST_CASE("/GlobalTagThrottler/TagLimit") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 0);
|
|
state StorageServerCollection storageServers(10, 5);
|
|
std::vector<Future<Void>> futures;
|
|
for (int i = 0; i < 2 * SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MAX_TAGS_TRACKED; ++i) {
|
|
Arena arena;
|
|
TransactionTag tag = makeString(8, arena);
|
|
deterministicRandom()->randomBytes(mutateString(tag), tag.size());
|
|
futures.push_back(runClient(&globalTagThrottler, &storageServers, tag, 1.0, 6.0, OpType::READ));
|
|
}
|
|
wait(timeout(waitForAll(futures), 60.0, Void()));
|
|
ASSERT_EQ(globalTagThrottler.tagsTracked(), SERVER_KNOBS->GLOBAL_TAG_THROTTLING_MAX_TAGS_TRACKED);
|
|
return Void();
|
|
}
|
|
|
|
// 9 storage servers can handle 100 pages/second each.
|
|
// 1 unhealthy storage server can only handle 1 page/second.
|
|
// Total quota is set to 100 pages/second.
|
|
// Client attempts 5 6-page transactions per second.
|
|
// Target rate adjusts to 100/6 transactions per second, ignoring the worst storage server.
|
|
// Then, a second storage server becomes unhealthy and can only handle 1 page/second.
|
|
// Target rate adjusts down to 10/6 transactions per second, because only one bad zone can be ignored.
|
|
TEST_CASE("/GlobalTagThrottler/IgnoreWorstZone") {
|
|
state GlobalTagThrottler globalTagThrottler(Database{}, UID{}, 1);
|
|
state StorageServerCollection storageServers(10, 100);
|
|
state TransactionTag testTag = "sampleTag1"_sr;
|
|
storageServers.setCapacity(0, 1);
|
|
ThrottleApi::TagQuotaValue tagQuotaValue;
|
|
tagQuotaValue.totalQuota = 100 * CLIENT_KNOBS->TAG_THROTTLING_PAGE_SIZE;
|
|
globalTagThrottler.setQuota(testTag, tagQuotaValue);
|
|
state Future<Void> client = runClient(&globalTagThrottler, &storageServers, testTag, 5.0, 6.0, OpType::READ);
|
|
state Future<Void> monitor = monitorActor(
|
|
&globalTagThrottler, [](auto& gtt) { return targetRateIsNear(gtt, "sampleTag1"_sr, 100.0 / 6.0); });
|
|
state Future<Void> updater = updateGlobalTagThrottler(&globalTagThrottler, &storageServers);
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
storageServers.setCapacity(1, 1);
|
|
monitor =
|
|
monitorActor(&globalTagThrottler, [](auto& gtt) { return targetRateIsNear(gtt, "sampleTag1"_sr, 10.0 / 6.0); });
|
|
wait(timeoutError(monitor || client || updater, 600.0));
|
|
return Void();
|
|
}
|