You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/table/merger.cc

428 lines
13 KiB

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree. An additional grant
// of patent rights can be found in the PATENTS file in the same directory.
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "table/merger.h"
#include <string>
#include <vector>
#include "db/dbformat.h"
#include "db/pinned_iterators_manager.h"
#include "rocksdb/comparator.h"
#include "rocksdb/iterator.h"
#include "rocksdb/options.h"
#include "table/internal_iterator.h"
#include "table/iter_heap.h"
#include "table/iterator_wrapper.h"
#include "util/arena.h"
#include "util/autovector.h"
#include "util/coding.h"
#include "util/heap.h"
#include "util/perf_context_imp.h"
#include "util/stop_watch.h"
#include "util/sync_point.h"
namespace rocksdb {
// Without anonymous namespace here, we fail the warning -Wmissing-prototypes
namespace {
typedef BinaryHeap<IteratorWrapper*, MaxIteratorComparator> MergerMaxIterHeap;
typedef BinaryHeap<IteratorWrapper*, MinIteratorComparator> MergerMinIterHeap;
} // namespace
const size_t kNumIterReserve = 4;
class MergingIterator : public InternalIterator {
public:
MergingIterator(const Comparator* comparator, InternalIterator** children,
int n, bool is_arena_mode,
const SliceTransform* const prefix_extractor)
: is_arena_mode_(is_arena_mode),
comparator_(comparator),
current_(nullptr),
direction_(kForward),
minHeap_(comparator_),
prefix_extractor_(prefix_extractor),
pinned_iters_mgr_(nullptr) {
children_.resize(n);
for (int i = 0; i < n; i++) {
children_[i].Set(children[i]);
}
for (auto& child : children_) {
if (child.Valid()) {
minHeap_.push(&child);
}
}
current_ = CurrentForward();
}
virtual void AddIterator(InternalIterator* iter) {
assert(direction_ == kForward);
children_.emplace_back(iter);
if (pinned_iters_mgr_) {
iter->SetPinnedItersMgr(pinned_iters_mgr_);
Introduce ReadOptions::pin_data (support zero copy for keys) Summary: This patch update the Iterator API to introduce new functions that allow users to keep the Slices returned by key() valid as long as the Iterator is not deleted ReadOptions::pin_data : If true keep loaded blocks in memory as long as the iterator is not deleted Iterator::IsKeyPinned() : If true, this mean that the Slice returned by key() is valid as long as the iterator is not deleted Also add a new option BlockBasedTableOptions::use_delta_encoding to allow users to disable delta_encoding if needed. Benchmark results (using https://phabricator.fb.com/P20083553) ``` // $ du -h /home/tec/local/normal.4K.Snappy/db10077 // 6.1G /home/tec/local/normal.4K.Snappy/db10077 // $ du -h /home/tec/local/zero.8K.LZ4/db10077 // 6.4G /home/tec/local/zero.8K.LZ4/db10077 // Benchmarks for shard db10077 // _build/opt/rocks/benchmark/rocks_copy_benchmark \ // --normal_db_path="/home/tec/local/normal.4K.Snappy/db10077" \ // --zero_db_path="/home/tec/local/zero.8K.LZ4/db10077" // First run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 1.73s 576.97m // BM_StringPiece 103.74% 1.67s 598.55m // ============================================================================ // Match rate : 1000000 / 1000000 // Second run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 611.99ms 1.63 // BM_StringPiece 203.76% 300.35ms 3.33 // ============================================================================ // Match rate : 1000000 / 1000000 ``` Test Plan: Unit tests Reviewers: sdong, igor, anthony, yhchiang, rven Reviewed By: rven Subscribers: dhruba, lovro, adsharma Differential Revision: https://reviews.facebook.net/D48999
9 years ago
}
auto new_wrapper = children_.back();
if (new_wrapper.Valid()) {
minHeap_.push(&new_wrapper);
current_ = CurrentForward();
}
}
virtual ~MergingIterator() {
for (auto& child : children_) {
child.DeleteIter(is_arena_mode_);
}
}
virtual bool Valid() const override { return (current_ != nullptr); }
virtual void SeekToFirst() override {
ClearHeaps();
for (auto& child : children_) {
child.SeekToFirst();
if (child.Valid()) {
minHeap_.push(&child);
}
}
direction_ = kForward;
current_ = CurrentForward();
}
virtual void SeekToLast() override {
ClearHeaps();
InitMaxHeap();
for (auto& child : children_) {
child.SeekToLast();
if (child.Valid()) {
maxHeap_->push(&child);
}
}
direction_ = kReverse;
current_ = CurrentReverse();
}
virtual void Seek(const Slice& target) override {
ClearHeaps();
for (auto& child : children_) {
{
PERF_TIMER_GUARD(seek_child_seek_time);
child.Seek(target);
}
PERF_COUNTER_ADD(seek_child_seek_count, 1);
if (child.Valid()) {
PERF_TIMER_GUARD(seek_min_heap_time);
minHeap_.push(&child);
}
}
direction_ = kForward;
{
PERF_TIMER_GUARD(seek_min_heap_time);
current_ = CurrentForward();
}
}
virtual void Next() override {
assert(Valid());
// Ensure that all children are positioned after key().
// If we are moving in the forward direction, it is already
// true for all of the non-current children since current_ is
// the smallest child and key() == current_->key().
if (direction_ != kForward) {
// Otherwise, advance the non-current children. We advance current_
// just after the if-block.
ClearHeaps();
for (auto& child : children_) {
if (&child != current_) {
if (prefix_extractor_ == nullptr) {
child.Seek(key());
fix bug in merge_iterator when data race happens Summary: core dump when run `./db_stress --max_background_compactions=1 --max_write_buffer_number=3 --sync=0 --reopen=20 --write_buffer_size=33554432 --delpercent=5 --log2_keys_per_lock=10 --block_size=16384 --allow_concurrent_memtable_write=1 --test_batches_snapshots=0 --max_bytes_for_level_base=67108864 --progress_reports=0 --mmap_read=1 --kill_prefix_blacklist=WritableFileWriter::Append,WritableFileWriter::WriteBuffered --writepercent=35 --disable_data_sync=0 --readpercent=50 --subcompactions=3 --ops_per_thread=20000000 --memtablerep=skip_list --prefix_size=0 --target_file_size_multiplier=1 --column_families=1 --db=/dev/shm/rocksdb/rocksdb_crashtest_whitebox --threads=32 --disable_wal=0 --open_files=500000 --destroy_db_initially=0 --target_file_size_base=16777216 --nooverwritepercent=1 --iterpercent=10 --max_key=100000000 --prefixpercent=0 --use_clock_cache=false --kill_random_test=189 --cache_size=1048576 --verify_checksum=1` Actually the relevant flag is `--threads`, data race when --thread > 1 cause problem. It is possible that multiple threads read/write memtable simultaneously. After one thread calls Prev(), another thread may insert a new key just between the current key and the key next, which may cause the assert(current_ == CurrentForward()) failure when the first thread calls Next() again if in prefix seek mode Test Plan: rerun db_stress with >1 thread / make all check -j64 Reviewers: sdong, andrewkr, IslamAbdelRahman Reviewed By: IslamAbdelRahman Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D62979
8 years ago
if (child.Valid() && comparator_->Equal(key(), child.key())) {
child.Next();
}
} else {
// only for prefix_seek_mode
// we should not call Seek() here
if (child.Valid()) {
child.Next();
} else {
child.SeekToFirst();
}
}
fix bug in merge_iterator when data race happens Summary: core dump when run `./db_stress --max_background_compactions=1 --max_write_buffer_number=3 --sync=0 --reopen=20 --write_buffer_size=33554432 --delpercent=5 --log2_keys_per_lock=10 --block_size=16384 --allow_concurrent_memtable_write=1 --test_batches_snapshots=0 --max_bytes_for_level_base=67108864 --progress_reports=0 --mmap_read=1 --kill_prefix_blacklist=WritableFileWriter::Append,WritableFileWriter::WriteBuffered --writepercent=35 --disable_data_sync=0 --readpercent=50 --subcompactions=3 --ops_per_thread=20000000 --memtablerep=skip_list --prefix_size=0 --target_file_size_multiplier=1 --column_families=1 --db=/dev/shm/rocksdb/rocksdb_crashtest_whitebox --threads=32 --disable_wal=0 --open_files=500000 --destroy_db_initially=0 --target_file_size_base=16777216 --nooverwritepercent=1 --iterpercent=10 --max_key=100000000 --prefixpercent=0 --use_clock_cache=false --kill_random_test=189 --cache_size=1048576 --verify_checksum=1` Actually the relevant flag is `--threads`, data race when --thread > 1 cause problem. It is possible that multiple threads read/write memtable simultaneously. After one thread calls Prev(), another thread may insert a new key just between the current key and the key next, which may cause the assert(current_ == CurrentForward()) failure when the first thread calls Next() again if in prefix seek mode Test Plan: rerun db_stress with >1 thread / make all check -j64 Reviewers: sdong, andrewkr, IslamAbdelRahman Reviewed By: IslamAbdelRahman Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D62979
8 years ago
// This condition is needed because it is possible that multiple
// threads read/write memtable simultaneously. After one thread
// calls Prev(), another thread may insert a new key just between
// the current key and the key next, which may cause the
// assert(current_ == CurrentForward()) failure when the first
// thread calls Next() again if in prefix seek mode
while (child.Valid() &&
comparator_->Compare(key(), child.key()) >= 0) {
child.Next();
}
}
if (child.Valid()) {
bool skip_iter =
prefix_extractor_ != nullptr &&
prefix_extractor_->InDomain(ExtractUserKey(child.key())) &&
prefix_extractor_->Transform(ExtractUserKey(child.key())) !=
Slice(*prefix_);
if (&child == current_ || !skip_iter) {
minHeap_.push(&child);
}
}
}
direction_ = kForward;
// The loop advanced all non-current children to be > key() so current_
// should still be strictly the smallest key.
assert(current_ == CurrentForward());
}
// For the heap modifications below to be correct, current_ must be the
// current top of the heap.
assert(current_ == CurrentForward());
// as the current points to the current record. move the iterator forward.
current_->Next();
if (current_->Valid()) {
// current is still valid after the Next() call above. Call
// replace_top() to restore the heap property. When the same child
// iterator yields a sequence of keys, this is cheap.
minHeap_.replace_top(current_);
} else {
// current stopped being valid, remove it from the heap.
minHeap_.pop();
}
current_ = CurrentForward();
}
virtual void Prev() override {
assert(Valid());
// Ensure that all children are positioned before key().
// If we are moving in the reverse direction, it is already
// true for all of the non-current children since current_ is
// the largest child and key() == current_->key().
if (direction_ != kReverse) {
// Otherwise, retreat the non-current children. We retreat current_
// just after the if-block.
ClearHeaps();
InitMaxHeap();
for (auto& child : children_) {
if (&child != current_) {
if (prefix_extractor_ == nullptr) {
child.Seek(key());
} else {
// only for prefix_seek_mode
// we should not call Seek() here
}
if (child.Valid()) {
// Child is at first entry >= key(). Step back one to be < key()
TEST_SYNC_POINT_CALLBACK("MergeIterator::Prev:BeforePrev", &child);
child.Prev();
} else {
// Child has no entries >= key(). Position at last entry.
TEST_SYNC_POINT("MergeIterator::Prev:BeforeSeekToLast");
child.SeekToLast();
}
fix bug in merge_iterator when data race happens Summary: core dump when run `./db_stress --max_background_compactions=1 --max_write_buffer_number=3 --sync=0 --reopen=20 --write_buffer_size=33554432 --delpercent=5 --log2_keys_per_lock=10 --block_size=16384 --allow_concurrent_memtable_write=1 --test_batches_snapshots=0 --max_bytes_for_level_base=67108864 --progress_reports=0 --mmap_read=1 --kill_prefix_blacklist=WritableFileWriter::Append,WritableFileWriter::WriteBuffered --writepercent=35 --disable_data_sync=0 --readpercent=50 --subcompactions=3 --ops_per_thread=20000000 --memtablerep=skip_list --prefix_size=0 --target_file_size_multiplier=1 --column_families=1 --db=/dev/shm/rocksdb/rocksdb_crashtest_whitebox --threads=32 --disable_wal=0 --open_files=500000 --destroy_db_initially=0 --target_file_size_base=16777216 --nooverwritepercent=1 --iterpercent=10 --max_key=100000000 --prefixpercent=0 --use_clock_cache=false --kill_random_test=189 --cache_size=1048576 --verify_checksum=1` Actually the relevant flag is `--threads`, data race when --thread > 1 cause problem. It is possible that multiple threads read/write memtable simultaneously. After one thread calls Prev(), another thread may insert a new key just between the current key and the key next, which may cause the assert(current_ == CurrentForward()) failure when the first thread calls Next() again if in prefix seek mode Test Plan: rerun db_stress with >1 thread / make all check -j64 Reviewers: sdong, andrewkr, IslamAbdelRahman Reviewed By: IslamAbdelRahman Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D62979
8 years ago
while (child.Valid() &&
comparator_->Compare(key(), child.key()) <= 0) {
child.Prev();
}
}
if (child.Valid()) {
bool skip_iter =
prefix_extractor_ != nullptr &&
prefix_extractor_->InDomain(ExtractUserKey(child.key())) &&
prefix_extractor_->Transform(ExtractUserKey(child.key())) !=
Slice(*prefix_);
if (&child == current_ || !skip_iter) {
maxHeap_->push(&child);
}
}
}
direction_ = kReverse;
// Note that we don't do assert(current_ == CurrentReverse()) here
// because it is possible to have some keys larger than the seek-key
// inserted between Seek() and SeekToLast(), which makes current_ not
// equal to CurrentReverse().
current_ = CurrentReverse();
}
// For the heap modifications below to be correct, current_ must be the
// current top of the heap.
assert(current_ == CurrentReverse());
current_->Prev();
if (current_->Valid()) {
// current is still valid after the Prev() call above. Call
// replace_top() to restore the heap property. When the same child
// iterator yields a sequence of keys, this is cheap.
maxHeap_->replace_top(current_);
} else {
// current stopped being valid, remove it from the heap.
maxHeap_->pop();
}
current_ = CurrentReverse();
}
virtual Slice key() const override {
assert(Valid());
return current_->key();
}
virtual Slice value() const override {
assert(Valid());
return current_->value();
}
virtual Status status() const override {
Status s;
for (auto& child : children_) {
s = child.status();
if (!s.ok()) {
break;
}
}
return s;
}
virtual void SetPinnedItersMgr(
PinnedIteratorsManager* pinned_iters_mgr) override {
pinned_iters_mgr_ = pinned_iters_mgr;
Introduce ReadOptions::pin_data (support zero copy for keys) Summary: This patch update the Iterator API to introduce new functions that allow users to keep the Slices returned by key() valid as long as the Iterator is not deleted ReadOptions::pin_data : If true keep loaded blocks in memory as long as the iterator is not deleted Iterator::IsKeyPinned() : If true, this mean that the Slice returned by key() is valid as long as the iterator is not deleted Also add a new option BlockBasedTableOptions::use_delta_encoding to allow users to disable delta_encoding if needed. Benchmark results (using https://phabricator.fb.com/P20083553) ``` // $ du -h /home/tec/local/normal.4K.Snappy/db10077 // 6.1G /home/tec/local/normal.4K.Snappy/db10077 // $ du -h /home/tec/local/zero.8K.LZ4/db10077 // 6.4G /home/tec/local/zero.8K.LZ4/db10077 // Benchmarks for shard db10077 // _build/opt/rocks/benchmark/rocks_copy_benchmark \ // --normal_db_path="/home/tec/local/normal.4K.Snappy/db10077" \ // --zero_db_path="/home/tec/local/zero.8K.LZ4/db10077" // First run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 1.73s 576.97m // BM_StringPiece 103.74% 1.67s 598.55m // ============================================================================ // Match rate : 1000000 / 1000000 // Second run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 611.99ms 1.63 // BM_StringPiece 203.76% 300.35ms 3.33 // ============================================================================ // Match rate : 1000000 / 1000000 ``` Test Plan: Unit tests Reviewers: sdong, igor, anthony, yhchiang, rven Reviewed By: rven Subscribers: dhruba, lovro, adsharma Differential Revision: https://reviews.facebook.net/D48999
9 years ago
for (auto& child : children_) {
child.SetPinnedItersMgr(pinned_iters_mgr);
Introduce ReadOptions::pin_data (support zero copy for keys) Summary: This patch update the Iterator API to introduce new functions that allow users to keep the Slices returned by key() valid as long as the Iterator is not deleted ReadOptions::pin_data : If true keep loaded blocks in memory as long as the iterator is not deleted Iterator::IsKeyPinned() : If true, this mean that the Slice returned by key() is valid as long as the iterator is not deleted Also add a new option BlockBasedTableOptions::use_delta_encoding to allow users to disable delta_encoding if needed. Benchmark results (using https://phabricator.fb.com/P20083553) ``` // $ du -h /home/tec/local/normal.4K.Snappy/db10077 // 6.1G /home/tec/local/normal.4K.Snappy/db10077 // $ du -h /home/tec/local/zero.8K.LZ4/db10077 // 6.4G /home/tec/local/zero.8K.LZ4/db10077 // Benchmarks for shard db10077 // _build/opt/rocks/benchmark/rocks_copy_benchmark \ // --normal_db_path="/home/tec/local/normal.4K.Snappy/db10077" \ // --zero_db_path="/home/tec/local/zero.8K.LZ4/db10077" // First run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 1.73s 576.97m // BM_StringPiece 103.74% 1.67s 598.55m // ============================================================================ // Match rate : 1000000 / 1000000 // Second run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 611.99ms 1.63 // BM_StringPiece 203.76% 300.35ms 3.33 // ============================================================================ // Match rate : 1000000 / 1000000 ``` Test Plan: Unit tests Reviewers: sdong, igor, anthony, yhchiang, rven Reviewed By: rven Subscribers: dhruba, lovro, adsharma Differential Revision: https://reviews.facebook.net/D48999
9 years ago
}
}
virtual bool IsKeyPinned() const override {
assert(Valid());
return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
current_->IsKeyPinned();
Introduce ReadOptions::pin_data (support zero copy for keys) Summary: This patch update the Iterator API to introduce new functions that allow users to keep the Slices returned by key() valid as long as the Iterator is not deleted ReadOptions::pin_data : If true keep loaded blocks in memory as long as the iterator is not deleted Iterator::IsKeyPinned() : If true, this mean that the Slice returned by key() is valid as long as the iterator is not deleted Also add a new option BlockBasedTableOptions::use_delta_encoding to allow users to disable delta_encoding if needed. Benchmark results (using https://phabricator.fb.com/P20083553) ``` // $ du -h /home/tec/local/normal.4K.Snappy/db10077 // 6.1G /home/tec/local/normal.4K.Snappy/db10077 // $ du -h /home/tec/local/zero.8K.LZ4/db10077 // 6.4G /home/tec/local/zero.8K.LZ4/db10077 // Benchmarks for shard db10077 // _build/opt/rocks/benchmark/rocks_copy_benchmark \ // --normal_db_path="/home/tec/local/normal.4K.Snappy/db10077" \ // --zero_db_path="/home/tec/local/zero.8K.LZ4/db10077" // First run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 1.73s 576.97m // BM_StringPiece 103.74% 1.67s 598.55m // ============================================================================ // Match rate : 1000000 / 1000000 // Second run // ============================================================================ // rocks/benchmark/RocksCopyBenchmark.cpp relative time/iter iters/s // ============================================================================ // BM_StringCopy 611.99ms 1.63 // BM_StringPiece 203.76% 300.35ms 3.33 // ============================================================================ // Match rate : 1000000 / 1000000 ``` Test Plan: Unit tests Reviewers: sdong, igor, anthony, yhchiang, rven Reviewed By: rven Subscribers: dhruba, lovro, adsharma Differential Revision: https://reviews.facebook.net/D48999
9 years ago
}
Introduce FullMergeV2 (eliminate memcpy from merge operators) Summary: This diff update the code to pin the merge operator operands while the merge operation is done, so that we can eliminate the memcpy cost, to do that we need a new public API for FullMerge that replace the std::deque<std::string> with std::vector<Slice> This diff is stacked on top of D56493 and D56511 In this diff we - Update FullMergeV2 arguments to be encapsulated in MergeOperationInput and MergeOperationOutput which will make it easier to add new arguments in the future - Replace std::deque<std::string> with std::vector<Slice> to pass operands - Replace MergeContext std::deque with std::vector (based on a simple benchmark I ran https://gist.github.com/IslamAbdelRahman/78fc86c9ab9f52b1df791e58943fb187) - Allow FullMergeV2 output to be an existing operand ``` [Everything in Memtable | 10K operands | 10 KB each | 1 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=10000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 0.607 micros/op 1648235 ops/sec; 16121.2 MB/s readseq : 0.478 micros/op 2091546 ops/sec; 20457.2 MB/s readseq : 0.252 micros/op 3972081 ops/sec; 38850.5 MB/s readseq : 0.237 micros/op 4218328 ops/sec; 41259.0 MB/s readseq : 0.247 micros/op 4043927 ops/sec; 39553.2 MB/s [master] readseq : 3.935 micros/op 254140 ops/sec; 2485.7 MB/s readseq : 3.722 micros/op 268657 ops/sec; 2627.7 MB/s readseq : 3.149 micros/op 317605 ops/sec; 3106.5 MB/s readseq : 3.125 micros/op 320024 ops/sec; 3130.1 MB/s readseq : 4.075 micros/op 245374 ops/sec; 2400.0 MB/s ``` ``` [Everything in Memtable | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="mergerandom,readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --merge_keys=1000 --num=10000 --disable_auto_compactions --value_size=10240 --write_buffer_size=1000000000 [FullMergeV2] readseq : 3.472 micros/op 288018 ops/sec; 2817.1 MB/s readseq : 2.304 micros/op 434027 ops/sec; 4245.2 MB/s readseq : 1.163 micros/op 859845 ops/sec; 8410.0 MB/s readseq : 1.192 micros/op 838926 ops/sec; 8205.4 MB/s readseq : 1.250 micros/op 800000 ops/sec; 7824.7 MB/s [master] readseq : 24.025 micros/op 41623 ops/sec; 407.1 MB/s readseq : 18.489 micros/op 54086 ops/sec; 529.0 MB/s readseq : 18.693 micros/op 53495 ops/sec; 523.2 MB/s readseq : 23.621 micros/op 42335 ops/sec; 414.1 MB/s readseq : 18.775 micros/op 53262 ops/sec; 521.0 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 1 operand per key] [FullMergeV2] $ DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions readseq : 14.741 micros/op 67837 ops/sec; 663.5 MB/s readseq : 1.029 micros/op 971446 ops/sec; 9501.6 MB/s readseq : 0.974 micros/op 1026229 ops/sec; 10037.4 MB/s readseq : 0.965 micros/op 1036080 ops/sec; 10133.8 MB/s readseq : 0.943 micros/op 1060657 ops/sec; 10374.2 MB/s [master] readseq : 16.735 micros/op 59755 ops/sec; 584.5 MB/s readseq : 3.029 micros/op 330151 ops/sec; 3229.2 MB/s readseq : 3.136 micros/op 318883 ops/sec; 3119.0 MB/s readseq : 3.065 micros/op 326245 ops/sec; 3191.0 MB/s readseq : 3.014 micros/op 331813 ops/sec; 3245.4 MB/s ``` ``` [Everything in Block cache | 10K operands | 10 KB each | 10 operand per key] DEBUG_LEVEL=0 make db_bench -j64 && ./db_bench --benchmarks="readseq,readseq,readseq,readseq,readseq" --merge_operator="max" --num=100000 --db="/dev/shm/merge-random-10-operands-10K-10KB" --cache_size=1000000000 --use_existing_db --disable_auto_compactions [FullMergeV2] readseq : 24.325 micros/op 41109 ops/sec; 402.1 MB/s readseq : 1.470 micros/op 680272 ops/sec; 6653.7 MB/s readseq : 1.231 micros/op 812347 ops/sec; 7945.5 MB/s readseq : 1.091 micros/op 916590 ops/sec; 8965.1 MB/s readseq : 1.109 micros/op 901713 ops/sec; 8819.6 MB/s [master] readseq : 27.257 micros/op 36687 ops/sec; 358.8 MB/s readseq : 4.443 micros/op 225073 ops/sec; 2201.4 MB/s readseq : 5.830 micros/op 171526 ops/sec; 1677.7 MB/s readseq : 4.173 micros/op 239635 ops/sec; 2343.8 MB/s readseq : 4.150 micros/op 240963 ops/sec; 2356.8 MB/s ``` Test Plan: COMPILE_WITH_ASAN=1 make check -j64 Reviewers: yhchiang, andrewkr, sdong Reviewed By: sdong Subscribers: lovro, andrewkr, dhruba Differential Revision: https://reviews.facebook.net/D57075
8 years ago
virtual bool IsValuePinned() const override {
assert(Valid());
return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
current_->IsValuePinned();
}
virtual void ResetPrefix(const Slice* prefix) override {
if (prefix == nullptr) {
prefix_.reset();
return;
}
if (!prefix_) {
prefix_.reset(new std::string);
}
*prefix_ = prefix->ToString();
}
private:
// Clears heaps for both directions, used when changing direction or seeking
void ClearHeaps();
// Ensures that maxHeap_ is initialized when starting to go in the reverse
// direction
void InitMaxHeap();
bool is_arena_mode_;
const Comparator* comparator_;
autovector<IteratorWrapper, kNumIterReserve> children_;
// Cached pointer to child iterator with the current key, or nullptr if no
// child iterators are valid. This is the top of minHeap_ or maxHeap_
// depending on the direction.
IteratorWrapper* current_;
// Which direction is the iterator moving?
enum Direction {
kForward,
kReverse
};
Direction direction_;
MergerMinIterHeap minHeap_;
// Max heap is used for reverse iteration, which is way less common than
// forward. Lazily initialize it to save memory.
std::unique_ptr<MergerMaxIterHeap> maxHeap_;
const SliceTransform* const prefix_extractor_;
PinnedIteratorsManager* pinned_iters_mgr_;
std::unique_ptr<std::string> prefix_;
IteratorWrapper* CurrentForward() const {
assert(direction_ == kForward);
return !minHeap_.empty() ? minHeap_.top() : nullptr;
Replace std::priority_queue in MergingIterator with custom heap Summary: While profiling compaction in our service I noticed a lot of CPU (~15% of compaction) being spent in MergingIterator and key comparison. Looking at the code I found MergingIterator was (understandably) using std::priority_queue for the multiway merge. Keys in our dataset include sequence numbers that increase with time. Adjacent keys in an L0 file are very likely to be adjacent in the full database. Consequently, compaction will often pick a chunk of rows from the same L0 file before switching to another one. It would be great to avoid the O(log K) operation per row while compacting. This diff replaces std::priority_queue with a custom binary heap implementation. It has a "replace top" operation that is cheap when the new top is the same as the old one (i.e. the priority of the top entry is decreased but it still stays on top). Test Plan: make check To test the effect on performance, I generated databases with data patterns that mimic what I describe in the summary (rows have a mostly increasing sequence number). I see a 10-15% CPU decrease for compaction (and a matching throughput improvement on tmpfs). The exact improvement depends on the number of L0 files and the amount of locality. Performance on randomly distributed keys seems on par with the old code. Reviewers: kailiu, sdong, igor Reviewed By: igor Subscribers: yoshinorim, dhruba, tnovak Differential Revision: https://reviews.facebook.net/D29133
10 years ago
}
IteratorWrapper* CurrentReverse() const {
assert(direction_ == kReverse);
assert(maxHeap_);
return !maxHeap_->empty() ? maxHeap_->top() : nullptr;
Replace std::priority_queue in MergingIterator with custom heap Summary: While profiling compaction in our service I noticed a lot of CPU (~15% of compaction) being spent in MergingIterator and key comparison. Looking at the code I found MergingIterator was (understandably) using std::priority_queue for the multiway merge. Keys in our dataset include sequence numbers that increase with time. Adjacent keys in an L0 file are very likely to be adjacent in the full database. Consequently, compaction will often pick a chunk of rows from the same L0 file before switching to another one. It would be great to avoid the O(log K) operation per row while compacting. This diff replaces std::priority_queue with a custom binary heap implementation. It has a "replace top" operation that is cheap when the new top is the same as the old one (i.e. the priority of the top entry is decreased but it still stays on top). Test Plan: make check To test the effect on performance, I generated databases with data patterns that mimic what I describe in the summary (rows have a mostly increasing sequence number). I see a 10-15% CPU decrease for compaction (and a matching throughput improvement on tmpfs). The exact improvement depends on the number of L0 files and the amount of locality. Performance on randomly distributed keys seems on par with the old code. Reviewers: kailiu, sdong, igor Reviewed By: igor Subscribers: yoshinorim, dhruba, tnovak Differential Revision: https://reviews.facebook.net/D29133
10 years ago
}
};
void MergingIterator::ClearHeaps() {
minHeap_.clear();
if (maxHeap_) {
maxHeap_->clear();
}
}
void MergingIterator::InitMaxHeap() {
if (!maxHeap_) {
maxHeap_.reset(new MergerMaxIterHeap(comparator_));
}
}
InternalIterator* NewMergingIterator(
const Comparator* cmp, InternalIterator** list, int n, Arena* arena,
const SliceTransform* const prefix_extractor) {
assert(n >= 0);
if (n == 0) {
return NewEmptyInternalIterator(arena);
} else if (n == 1) {
return list[0];
} else {
if (arena == nullptr) {
return new MergingIterator(cmp, list, n, false, prefix_extractor);
} else {
auto mem = arena->AllocateAligned(sizeof(MergingIterator));
return new (mem) MergingIterator(cmp, list, n, true, prefix_extractor);
}
}
}
MergeIteratorBuilder::MergeIteratorBuilder(
const Comparator* comparator, Arena* a,
const SliceTransform* const prefix_extractor)
: first_iter(nullptr), use_merging_iter(false), arena(a) {
auto mem = arena->AllocateAligned(sizeof(MergingIterator));
merge_iter =
new (mem) MergingIterator(comparator, nullptr, 0, true, prefix_extractor);
}
void MergeIteratorBuilder::AddIterator(InternalIterator* iter) {
if (!use_merging_iter && first_iter != nullptr) {
merge_iter->AddIterator(first_iter);
use_merging_iter = true;
}
if (use_merging_iter) {
merge_iter->AddIterator(iter);
} else {
first_iter = iter;
}
}
InternalIterator* MergeIteratorBuilder::Finish() {
if (!use_merging_iter) {
return first_iter;
} else {
auto ret = merge_iter;
merge_iter = nullptr;
return ret;
}
}
} // namespace rocksdb