2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
2013-10-16 23:59:46 +02:00
|
|
|
//
|
2011-06-22 20:45:39 +02:00
|
|
|
// 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 "db/version_set.h"
|
2018-06-28 21:16:10 +02:00
|
|
|
#include "db/log_writer.h"
|
|
|
|
#include "table/mock_table.h"
|
2011-06-22 20:45:39 +02:00
|
|
|
#include "util/logging.h"
|
2018-07-14 02:34:54 +02:00
|
|
|
#include "util/string_util.h"
|
2011-06-22 20:45:39 +02:00
|
|
|
#include "util/testharness.h"
|
|
|
|
#include "util/testutil.h"
|
|
|
|
|
2013-10-04 06:49:15 +02:00
|
|
|
namespace rocksdb {
|
2011-06-22 20:45:39 +02:00
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
class GenerateLevelFilesBriefTest : public testing::Test {
|
2011-06-22 20:45:39 +02:00
|
|
|
public:
|
|
|
|
std::vector<FileMetaData*> files_;
|
2014-10-28 18:03:13 +01:00
|
|
|
LevelFilesBrief file_level_;
|
2014-07-11 21:52:41 +02:00
|
|
|
Arena arena_;
|
2011-10-06 01:30:28 +02:00
|
|
|
|
2014-10-28 18:03:13 +01:00
|
|
|
GenerateLevelFilesBriefTest() { }
|
2011-06-22 20:45:39 +02:00
|
|
|
|
2014-10-28 18:03:13 +01:00
|
|
|
~GenerateLevelFilesBriefTest() {
|
2015-12-16 00:26:20 +01:00
|
|
|
for (size_t i = 0; i < files_.size(); i++) {
|
2011-06-22 20:45:39 +02:00
|
|
|
delete files_[i];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-07-15 02:20:57 +02:00
|
|
|
void Add(const char* smallest, const char* largest,
|
|
|
|
SequenceNumber smallest_seq = 100,
|
|
|
|
SequenceNumber largest_seq = 100) {
|
2011-06-22 20:45:39 +02:00
|
|
|
FileMetaData* f = new FileMetaData;
|
2014-07-02 18:54:20 +02:00
|
|
|
f->fd = FileDescriptor(files_.size() + 1, 0, 0);
|
2011-07-15 02:20:57 +02:00
|
|
|
f->smallest = InternalKey(smallest, smallest_seq, kTypeValue);
|
|
|
|
f->largest = InternalKey(largest, largest_seq, kTypeValue);
|
2011-06-22 20:45:39 +02:00
|
|
|
files_.push_back(f);
|
|
|
|
}
|
|
|
|
|
2014-07-11 21:52:41 +02:00
|
|
|
int Compare() {
|
|
|
|
int diff = 0;
|
|
|
|
for (size_t i = 0; i < files_.size(); i++) {
|
|
|
|
if (file_level_.files[i].fd.GetNumber() != files_[i]->fd.GetNumber()) {
|
|
|
|
diff++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return diff;
|
2011-06-22 20:45:39 +02:00
|
|
|
}
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
};
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(GenerateLevelFilesBriefTest, Empty) {
|
2014-10-28 18:03:13 +01:00
|
|
|
DoGenerateLevelFilesBrief(&file_level_, files_, &arena_);
|
2014-07-20 16:56:40 +02:00
|
|
|
ASSERT_EQ(0u, file_level_.num_files);
|
2014-07-11 21:52:41 +02:00
|
|
|
ASSERT_EQ(0, Compare());
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(GenerateLevelFilesBriefTest, Single) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
Add("p", "q");
|
2014-10-28 18:03:13 +01:00
|
|
|
DoGenerateLevelFilesBrief(&file_level_, files_, &arena_);
|
2014-07-20 16:56:40 +02:00
|
|
|
ASSERT_EQ(1u, file_level_.num_files);
|
2014-07-11 21:52:41 +02:00
|
|
|
ASSERT_EQ(0, Compare());
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(GenerateLevelFilesBriefTest, Multiple) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
Add("150", "200");
|
|
|
|
Add("200", "250");
|
|
|
|
Add("300", "350");
|
|
|
|
Add("400", "450");
|
2014-10-28 18:03:13 +01:00
|
|
|
DoGenerateLevelFilesBrief(&file_level_, files_, &arena_);
|
2014-07-20 16:56:40 +02:00
|
|
|
ASSERT_EQ(4u, file_level_.num_files);
|
2014-07-11 21:52:41 +02:00
|
|
|
ASSERT_EQ(0, Compare());
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
}
|
|
|
|
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
class CountingLogger : public Logger {
|
|
|
|
public:
|
|
|
|
CountingLogger() : log_count(0) {}
|
|
|
|
using Logger::Logv;
|
2018-03-05 22:08:17 +01:00
|
|
|
virtual void Logv(const char* /*format*/, va_list /*ap*/) override {
|
|
|
|
log_count++;
|
|
|
|
}
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
int log_count;
|
|
|
|
};
|
|
|
|
|
|
|
|
Options GetOptionsWithNumLevels(int num_levels,
|
|
|
|
std::shared_ptr<CountingLogger> logger) {
|
|
|
|
Options opt;
|
|
|
|
opt.num_levels = num_levels;
|
|
|
|
opt.info_log = logger;
|
|
|
|
return opt;
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
class VersionStorageInfoTest : public testing::Test {
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
public:
|
|
|
|
const Comparator* ucmp_;
|
|
|
|
InternalKeyComparator icmp_;
|
|
|
|
std::shared_ptr<CountingLogger> logger_;
|
|
|
|
Options options_;
|
|
|
|
ImmutableCFOptions ioptions_;
|
|
|
|
MutableCFOptions mutable_cf_options_;
|
|
|
|
VersionStorageInfo vstorage_;
|
|
|
|
|
|
|
|
InternalKey GetInternalKey(const char* ukey,
|
|
|
|
SequenceNumber smallest_seq = 100) {
|
|
|
|
return InternalKey(ukey, smallest_seq, kTypeValue);
|
|
|
|
}
|
|
|
|
|
|
|
|
VersionStorageInfoTest()
|
|
|
|
: ucmp_(BytewiseComparator()),
|
|
|
|
icmp_(ucmp_),
|
|
|
|
logger_(new CountingLogger()),
|
|
|
|
options_(GetOptionsWithNumLevels(6, logger_)),
|
|
|
|
ioptions_(options_),
|
2016-09-14 06:11:59 +02:00
|
|
|
mutable_cf_options_(options_),
|
2016-10-08 02:21:45 +02:00
|
|
|
vstorage_(&icmp_, ucmp_, 6, kCompactionStyleLevel, nullptr, false) {}
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
|
|
|
|
~VersionStorageInfoTest() {
|
|
|
|
for (int i = 0; i < vstorage_.num_levels(); i++) {
|
|
|
|
for (auto* f : vstorage_.LevelFiles(i)) {
|
|
|
|
if (--f->refs == 0) {
|
|
|
|
delete f;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void Add(int level, uint32_t file_number, const char* smallest,
|
|
|
|
const char* largest, uint64_t file_size = 0) {
|
|
|
|
assert(level < vstorage_.num_levels());
|
|
|
|
FileMetaData* f = new FileMetaData;
|
|
|
|
f->fd = FileDescriptor(file_number, 0, file_size);
|
|
|
|
f->smallest = GetInternalKey(smallest, 0);
|
|
|
|
f->largest = GetInternalKey(largest, 0);
|
|
|
|
f->compensated_file_size = file_size;
|
|
|
|
f->refs = 0;
|
|
|
|
f->num_entries = 0;
|
|
|
|
f->num_deletions = 0;
|
|
|
|
vstorage_.AddFile(level, f);
|
|
|
|
}
|
2018-07-14 02:34:54 +02:00
|
|
|
|
|
|
|
void Add(int level, uint32_t file_number, const InternalKey& smallest,
|
|
|
|
const InternalKey& largest, uint64_t file_size = 0) {
|
|
|
|
assert(level < vstorage_.num_levels());
|
|
|
|
FileMetaData* f = new FileMetaData;
|
|
|
|
f->fd = FileDescriptor(file_number, 0, file_size);
|
|
|
|
f->smallest = smallest;
|
|
|
|
f->largest = largest;
|
|
|
|
f->compensated_file_size = file_size;
|
|
|
|
f->refs = 0;
|
|
|
|
f->num_entries = 0;
|
|
|
|
f->num_deletions = 0;
|
|
|
|
vstorage_.AddFile(level, f);
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string GetOverlappingFiles(int level, const InternalKey& begin,
|
|
|
|
const InternalKey& end) {
|
|
|
|
std::vector<FileMetaData*> inputs;
|
|
|
|
vstorage_.GetOverlappingInputs(level, &begin, &end, &inputs);
|
|
|
|
|
|
|
|
std::string result;
|
|
|
|
for (size_t i = 0; i < inputs.size(); ++i) {
|
|
|
|
if (i > 0) {
|
|
|
|
result += ",";
|
|
|
|
}
|
|
|
|
AppendNumberTo(&result, inputs[i]->fd.GetNumber());
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
};
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(VersionStorageInfoTest, MaxBytesForLevelStatic) {
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ioptions_.level_compaction_dynamic_level_bytes = false;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_base = 10;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_multiplier = 5;
|
|
|
|
Add(4, 100U, "1", "2");
|
|
|
|
Add(5, 101U, "1", "2");
|
|
|
|
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(1), 10U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(2), 50U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(3), 250U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(4), 1250U);
|
|
|
|
|
|
|
|
ASSERT_EQ(0, logger_->log_count);
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(VersionStorageInfoTest, MaxBytesForLevelDynamic) {
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ioptions_.level_compaction_dynamic_level_bytes = true;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_base = 1000;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_multiplier = 5;
|
|
|
|
Add(5, 1U, "1", "2", 500U);
|
|
|
|
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(0, logger_->log_count);
|
|
|
|
ASSERT_EQ(vstorage_.base_level(), 5);
|
|
|
|
|
|
|
|
Add(5, 2U, "3", "4", 550U);
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(0, logger_->log_count);
|
2017-05-05 03:03:22 +02:00
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(4), 1000U);
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ASSERT_EQ(vstorage_.base_level(), 4);
|
|
|
|
|
|
|
|
Add(4, 3U, "3", "4", 550U);
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(0, logger_->log_count);
|
2017-05-05 03:03:22 +02:00
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(4), 1000U);
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ASSERT_EQ(vstorage_.base_level(), 4);
|
|
|
|
|
|
|
|
Add(3, 4U, "3", "4", 250U);
|
|
|
|
Add(3, 5U, "5", "7", 300U);
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(1, logger_->log_count);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(4), 1005U);
|
2017-05-05 03:03:22 +02:00
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(3), 1000U);
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ASSERT_EQ(vstorage_.base_level(), 3);
|
|
|
|
|
|
|
|
Add(1, 6U, "3", "4", 5U);
|
|
|
|
Add(1, 7U, "8", "9", 5U);
|
|
|
|
logger_->log_count = 0;
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(1, logger_->log_count);
|
|
|
|
ASSERT_GT(vstorage_.MaxBytesForLevel(4), 1005U);
|
|
|
|
ASSERT_GT(vstorage_.MaxBytesForLevel(3), 1005U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(2), 1005U);
|
2017-05-05 03:03:22 +02:00
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(1), 1000U);
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ASSERT_EQ(vstorage_.base_level(), 1);
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(VersionStorageInfoTest, MaxBytesForLevelDynamicLotsOfData) {
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
ioptions_.level_compaction_dynamic_level_bytes = true;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_base = 100;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_multiplier = 2;
|
|
|
|
Add(0, 1U, "1", "2", 50U);
|
|
|
|
Add(1, 2U, "1", "2", 50U);
|
|
|
|
Add(2, 3U, "1", "2", 500U);
|
|
|
|
Add(3, 4U, "1", "2", 500U);
|
|
|
|
Add(4, 5U, "1", "2", 1700U);
|
|
|
|
Add(5, 6U, "1", "2", 500U);
|
|
|
|
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(4), 800U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(3), 400U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(2), 200U);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(1), 100U);
|
|
|
|
ASSERT_EQ(vstorage_.base_level(), 1);
|
|
|
|
ASSERT_EQ(0, logger_->log_count);
|
2015-04-03 07:24:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(VersionStorageInfoTest, MaxBytesForLevelDynamicLargeLevel) {
|
|
|
|
uint64_t kOneGB = 1000U * 1000U * 1000U;
|
|
|
|
ioptions_.level_compaction_dynamic_level_bytes = true;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_base = 10U * kOneGB;
|
|
|
|
mutable_cf_options_.max_bytes_for_level_multiplier = 10;
|
|
|
|
Add(0, 1U, "1", "2", 50U);
|
|
|
|
Add(3, 4U, "1", "2", 32U * kOneGB);
|
|
|
|
Add(4, 5U, "1", "2", 500U * kOneGB);
|
|
|
|
Add(5, 6U, "1", "2", 3000U * kOneGB);
|
|
|
|
|
|
|
|
vstorage_.CalculateBaseBytes(ioptions_, mutable_cf_options_);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(5), 3000U * kOneGB);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(4), 300U * kOneGB);
|
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(3), 30U * kOneGB);
|
2017-05-05 03:03:22 +02:00
|
|
|
ASSERT_EQ(vstorage_.MaxBytesForLevel(2), 10U * kOneGB);
|
2015-04-03 07:24:50 +02:00
|
|
|
ASSERT_EQ(vstorage_.base_level(), 2);
|
|
|
|
ASSERT_EQ(0, logger_->log_count);
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 20:44:17 +01:00
|
|
|
}
|
|
|
|
|
2015-07-22 06:33:20 +02:00
|
|
|
TEST_F(VersionStorageInfoTest, EstimateLiveDataSize) {
|
|
|
|
// Test whether the overlaps are detected as expected
|
|
|
|
Add(1, 1U, "4", "7", 1U); // Perfect overlap with last level
|
|
|
|
Add(2, 2U, "3", "5", 1U); // Partial overlap with last level
|
|
|
|
Add(2, 3U, "6", "8", 1U); // Partial overlap with last level
|
|
|
|
Add(3, 4U, "1", "9", 1U); // Contains range of last level
|
|
|
|
Add(4, 5U, "4", "5", 1U); // Inside range of last level
|
|
|
|
Add(4, 5U, "6", "7", 1U); // Inside range of last level
|
|
|
|
Add(5, 6U, "4", "7", 10U);
|
|
|
|
ASSERT_EQ(10U, vstorage_.EstimateLiveDataSize());
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(VersionStorageInfoTest, EstimateLiveDataSize2) {
|
|
|
|
Add(0, 1U, "9", "9", 1U); // Level 0 is not ordered
|
|
|
|
Add(0, 1U, "5", "6", 1U); // Ignored because of [5,6] in l1
|
|
|
|
Add(1, 1U, "1", "2", 1U); // Ignored because of [2,3] in l2
|
|
|
|
Add(1, 2U, "3", "4", 1U); // Ignored because of [2,3] in l2
|
|
|
|
Add(1, 3U, "5", "6", 1U);
|
|
|
|
Add(2, 4U, "2", "3", 1U);
|
|
|
|
Add(3, 5U, "7", "8", 1U);
|
|
|
|
ASSERT_EQ(4U, vstorage_.EstimateLiveDataSize());
|
|
|
|
}
|
|
|
|
|
2018-07-14 02:34:54 +02:00
|
|
|
TEST_F(VersionStorageInfoTest, GetOverlappingInputs) {
|
|
|
|
// Two files that overlap at the range deletion tombstone sentinel.
|
|
|
|
Add(1, 1U, {"a", 0, kTypeValue}, {"b", kMaxSequenceNumber, kTypeRangeDeletion}, 1);
|
|
|
|
Add(1, 2U, {"b", 0, kTypeValue}, {"c", 0, kTypeValue}, 1);
|
|
|
|
// Two files that overlap at the same user key.
|
|
|
|
Add(1, 3U, {"d", 0, kTypeValue}, {"e", kMaxSequenceNumber, kTypeValue}, 1);
|
|
|
|
Add(1, 4U, {"e", 0, kTypeValue}, {"f", 0, kTypeValue}, 1);
|
|
|
|
// Two files that do not overlap.
|
|
|
|
Add(1, 5U, {"g", 0, kTypeValue}, {"h", 0, kTypeValue}, 1);
|
|
|
|
Add(1, 6U, {"i", 0, kTypeValue}, {"j", 0, kTypeValue}, 1);
|
|
|
|
vstorage_.UpdateNumNonEmptyLevels();
|
|
|
|
vstorage_.GenerateLevelFilesBrief();
|
|
|
|
|
|
|
|
ASSERT_EQ("1,2", GetOverlappingFiles(
|
|
|
|
1, {"a", 0, kTypeValue}, {"b", 0, kTypeValue}));
|
|
|
|
ASSERT_EQ("1", GetOverlappingFiles(
|
|
|
|
1, {"a", 0, kTypeValue}, {"b", kMaxSequenceNumber, kTypeRangeDeletion}));
|
|
|
|
ASSERT_EQ("2", GetOverlappingFiles(
|
|
|
|
1, {"b", kMaxSequenceNumber, kTypeValue}, {"c", 0, kTypeValue}));
|
|
|
|
ASSERT_EQ("3,4", GetOverlappingFiles(
|
|
|
|
1, {"d", 0, kTypeValue}, {"e", 0, kTypeValue}));
|
|
|
|
ASSERT_EQ("3", GetOverlappingFiles(
|
|
|
|
1, {"d", 0, kTypeValue}, {"e", kMaxSequenceNumber, kTypeRangeDeletion}));
|
|
|
|
ASSERT_EQ("3,4", GetOverlappingFiles(
|
|
|
|
1, {"e", kMaxSequenceNumber, kTypeValue}, {"f", 0, kTypeValue}));
|
|
|
|
ASSERT_EQ("3,4", GetOverlappingFiles(
|
|
|
|
1, {"e", 0, kTypeValue}, {"f", 0, kTypeValue}));
|
|
|
|
ASSERT_EQ("5", GetOverlappingFiles(
|
|
|
|
1, {"g", 0, kTypeValue}, {"h", 0, kTypeValue}));
|
|
|
|
ASSERT_EQ("6", GetOverlappingFiles(
|
|
|
|
1, {"i", 0, kTypeValue}, {"j", 0, kTypeValue}));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
class FindLevelFileTest : public testing::Test {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
public:
|
2014-10-28 18:03:13 +01:00
|
|
|
LevelFilesBrief file_level_;
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
bool disjoint_sorted_files_;
|
|
|
|
Arena arena_;
|
|
|
|
|
|
|
|
FindLevelFileTest() : disjoint_sorted_files_(true) { }
|
|
|
|
|
|
|
|
~FindLevelFileTest() {
|
|
|
|
}
|
|
|
|
|
|
|
|
void LevelFileInit(size_t num = 0) {
|
|
|
|
char* mem = arena_.AllocateAligned(num * sizeof(FdWithKeyRange));
|
2014-07-11 21:52:41 +02:00
|
|
|
file_level_.files = new (mem)FdWithKeyRange[num];
|
|
|
|
file_level_.num_files = 0;
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
void Add(const char* smallest, const char* largest,
|
|
|
|
SequenceNumber smallest_seq = 100,
|
|
|
|
SequenceNumber largest_seq = 100) {
|
|
|
|
InternalKey smallest_key = InternalKey(smallest, smallest_seq, kTypeValue);
|
|
|
|
InternalKey largest_key = InternalKey(largest, largest_seq, kTypeValue);
|
|
|
|
|
|
|
|
Slice smallest_slice = smallest_key.Encode();
|
|
|
|
Slice largest_slice = largest_key.Encode();
|
|
|
|
|
|
|
|
char* mem = arena_.AllocateAligned(
|
|
|
|
smallest_slice.size() + largest_slice.size());
|
|
|
|
memcpy(mem, smallest_slice.data(), smallest_slice.size());
|
|
|
|
memcpy(mem + smallest_slice.size(), largest_slice.data(),
|
|
|
|
largest_slice.size());
|
|
|
|
|
2014-07-11 21:52:41 +02:00
|
|
|
// add to file_level_
|
|
|
|
size_t num = file_level_.num_files;
|
|
|
|
auto& file = file_level_.files[num];
|
2014-07-10 08:40:03 +02:00
|
|
|
file.fd = FileDescriptor(num + 1, 0, 0);
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
file.smallest_key = Slice(mem, smallest_slice.size());
|
|
|
|
file.largest_key = Slice(mem + smallest_slice.size(),
|
|
|
|
largest_slice.size());
|
2014-07-11 21:52:41 +02:00
|
|
|
file_level_.num_files++;
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
int Find(const char* key) {
|
|
|
|
InternalKey target(key, 100, kTypeValue);
|
|
|
|
InternalKeyComparator cmp(BytewiseComparator());
|
2014-07-11 21:52:41 +02:00
|
|
|
return FindFile(cmp, file_level_, target.Encode());
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
}
|
2011-06-22 20:45:39 +02:00
|
|
|
|
|
|
|
bool Overlaps(const char* smallest, const char* largest) {
|
|
|
|
InternalKeyComparator cmp(BytewiseComparator());
|
2013-03-01 03:04:58 +01:00
|
|
|
Slice s(smallest != nullptr ? smallest : "");
|
|
|
|
Slice l(largest != nullptr ? largest : "");
|
2014-07-11 21:52:41 +02:00
|
|
|
return SomeFileOverlapsRange(cmp, disjoint_sorted_files_, file_level_,
|
2013-03-01 03:04:58 +01:00
|
|
|
(smallest != nullptr ? &s : nullptr),
|
|
|
|
(largest != nullptr ? &l : nullptr));
|
2011-06-22 20:45:39 +02:00
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(FindLevelFileTest, LevelEmpty) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
LevelFileInit(0);
|
|
|
|
|
2011-06-22 20:45:39 +02:00
|
|
|
ASSERT_EQ(0, Find("foo"));
|
|
|
|
ASSERT_TRUE(! Overlaps("a", "z"));
|
2013-03-01 03:04:58 +01:00
|
|
|
ASSERT_TRUE(! Overlaps(nullptr, "z"));
|
|
|
|
ASSERT_TRUE(! Overlaps("a", nullptr));
|
|
|
|
ASSERT_TRUE(! Overlaps(nullptr, nullptr));
|
2011-06-22 20:45:39 +02:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(FindLevelFileTest, LevelSingle) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
LevelFileInit(1);
|
|
|
|
|
2011-06-22 20:45:39 +02:00
|
|
|
Add("p", "q");
|
|
|
|
ASSERT_EQ(0, Find("a"));
|
|
|
|
ASSERT_EQ(0, Find("p"));
|
|
|
|
ASSERT_EQ(0, Find("p1"));
|
|
|
|
ASSERT_EQ(0, Find("q"));
|
|
|
|
ASSERT_EQ(1, Find("q1"));
|
|
|
|
ASSERT_EQ(1, Find("z"));
|
|
|
|
|
|
|
|
ASSERT_TRUE(! Overlaps("a", "b"));
|
|
|
|
ASSERT_TRUE(! Overlaps("z1", "z2"));
|
|
|
|
ASSERT_TRUE(Overlaps("a", "p"));
|
|
|
|
ASSERT_TRUE(Overlaps("a", "q"));
|
|
|
|
ASSERT_TRUE(Overlaps("a", "z"));
|
|
|
|
ASSERT_TRUE(Overlaps("p", "p1"));
|
|
|
|
ASSERT_TRUE(Overlaps("p", "q"));
|
|
|
|
ASSERT_TRUE(Overlaps("p", "z"));
|
|
|
|
ASSERT_TRUE(Overlaps("p1", "p2"));
|
|
|
|
ASSERT_TRUE(Overlaps("p1", "z"));
|
|
|
|
ASSERT_TRUE(Overlaps("q", "q"));
|
|
|
|
ASSERT_TRUE(Overlaps("q", "q1"));
|
2011-10-06 01:30:28 +02:00
|
|
|
|
2013-03-01 03:04:58 +01:00
|
|
|
ASSERT_TRUE(! Overlaps(nullptr, "j"));
|
|
|
|
ASSERT_TRUE(! Overlaps("r", nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "p"));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "p1"));
|
|
|
|
ASSERT_TRUE(Overlaps("q", nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, nullptr));
|
2011-06-22 20:45:39 +02:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(FindLevelFileTest, LevelMultiple) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
LevelFileInit(4);
|
2011-06-22 20:45:39 +02:00
|
|
|
|
|
|
|
Add("150", "200");
|
|
|
|
Add("200", "250");
|
|
|
|
Add("300", "350");
|
|
|
|
Add("400", "450");
|
|
|
|
ASSERT_EQ(0, Find("100"));
|
|
|
|
ASSERT_EQ(0, Find("150"));
|
|
|
|
ASSERT_EQ(0, Find("151"));
|
|
|
|
ASSERT_EQ(0, Find("199"));
|
|
|
|
ASSERT_EQ(0, Find("200"));
|
|
|
|
ASSERT_EQ(1, Find("201"));
|
|
|
|
ASSERT_EQ(1, Find("249"));
|
|
|
|
ASSERT_EQ(1, Find("250"));
|
|
|
|
ASSERT_EQ(2, Find("251"));
|
|
|
|
ASSERT_EQ(2, Find("299"));
|
|
|
|
ASSERT_EQ(2, Find("300"));
|
|
|
|
ASSERT_EQ(2, Find("349"));
|
|
|
|
ASSERT_EQ(2, Find("350"));
|
|
|
|
ASSERT_EQ(3, Find("351"));
|
|
|
|
ASSERT_EQ(3, Find("400"));
|
|
|
|
ASSERT_EQ(3, Find("450"));
|
|
|
|
ASSERT_EQ(4, Find("451"));
|
|
|
|
|
|
|
|
ASSERT_TRUE(! Overlaps("100", "149"));
|
|
|
|
ASSERT_TRUE(! Overlaps("251", "299"));
|
|
|
|
ASSERT_TRUE(! Overlaps("451", "500"));
|
|
|
|
ASSERT_TRUE(! Overlaps("351", "399"));
|
|
|
|
|
|
|
|
ASSERT_TRUE(Overlaps("100", "150"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "200"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "300"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "400"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "500"));
|
|
|
|
ASSERT_TRUE(Overlaps("375", "400"));
|
|
|
|
ASSERT_TRUE(Overlaps("450", "450"));
|
|
|
|
ASSERT_TRUE(Overlaps("450", "500"));
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(FindLevelFileTest, LevelMultipleNullBoundaries) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
LevelFileInit(4);
|
|
|
|
|
2011-10-06 01:30:28 +02:00
|
|
|
Add("150", "200");
|
|
|
|
Add("200", "250");
|
|
|
|
Add("300", "350");
|
|
|
|
Add("400", "450");
|
2013-03-01 03:04:58 +01:00
|
|
|
ASSERT_TRUE(! Overlaps(nullptr, "149"));
|
|
|
|
ASSERT_TRUE(! Overlaps("451", nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "150"));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "199"));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "200"));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "201"));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "400"));
|
|
|
|
ASSERT_TRUE(Overlaps(nullptr, "800"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps("200", nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps("449", nullptr));
|
|
|
|
ASSERT_TRUE(Overlaps("450", nullptr));
|
2011-10-06 01:30:28 +02:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(FindLevelFileTest, LevelOverlapSequenceChecks) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
LevelFileInit(1);
|
|
|
|
|
2011-07-15 02:20:57 +02:00
|
|
|
Add("200", "200", 5000, 3000);
|
|
|
|
ASSERT_TRUE(! Overlaps("199", "199"));
|
|
|
|
ASSERT_TRUE(! Overlaps("201", "300"));
|
|
|
|
ASSERT_TRUE(Overlaps("200", "200"));
|
|
|
|
ASSERT_TRUE(Overlaps("190", "200"));
|
|
|
|
ASSERT_TRUE(Overlaps("200", "210"));
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(FindLevelFileTest, LevelOverlappingFiles) {
|
create compressed_levels_ in Version, allocate its space using arena. Make Version::Get, Version::FindFile faster
Summary:
Define CompressedFileMetaData that just contains fd, smallest_slice, largest_slice. Create compressed_levels_ in Version, the space is allocated using arena
Thus increase the file meta data locality, speed up "Get" and "FindFile"
benchmark with in-memory tmpfs, could have 4% improvement under "random read" and 2% improvement under "read while writing"
benchmark command:
./db_bench --db=/mnt/db/rocksdb --num_levels=6 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --block_size=4096 --cache_size=17179869184 --cache_numshardbits=6 --compression_type=none --compression_ratio=1 --min_level_to_compress=-1 --disable_seek_compaction=1 --hard_rate_limit=2 --write_buffer_size=134217728 --max_write_buffer_number=2 --level0_file_num_compaction_trigger=8 --target_file_size_base=33554432 --max_bytes_for_level_base=1073741824 --disable_wal=0 --sync=0 --disable_data_sync=1 --verify_checksum=1 --delete_obsolete_files_period_micros=314572800 --max_grandparent_overlap_factor=10 --max_background_compactions=4 --max_background_flushes=0 --level0_slowdown_writes_trigger=16 --level0_stop_writes_trigger=24 --statistics=0 --stats_per_interval=0 --stats_interval=1048576 --histogram=0 --use_plain_table=1 --open_files=-1 --mmap_read=1 --mmap_write=0 --memtablerep=prefix_hash --bloom_bits=10 --bloom_locality=1 --perf_level=0 --benchmarks=readwhilewriting,readwhilewriting,readwhilewriting --use_existing_db=1 --num=52428800 --threads=1 —writes_per_second=81920
Read Random:
From 1.8363 ms/op, improve to 1.7587 ms/op.
Read while writing:
From 2.985 ms/op, improve to 2.924 ms/op.
Test Plan:
make all check
Reviewers: ljin, haobo, yhchiang, sdong
Reviewed By: sdong
Subscribers: dhruba, igor
Differential Revision: https://reviews.facebook.net/D19419
2014-07-10 07:14:39 +02:00
|
|
|
LevelFileInit(2);
|
|
|
|
|
2011-10-06 01:30:28 +02:00
|
|
|
Add("150", "600");
|
|
|
|
Add("400", "500");
|
|
|
|
disjoint_sorted_files_ = false;
|
|
|
|
ASSERT_TRUE(! Overlaps("100", "149"));
|
|
|
|
ASSERT_TRUE(! Overlaps("601", "700"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "150"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "200"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "300"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "400"));
|
|
|
|
ASSERT_TRUE(Overlaps("100", "500"));
|
|
|
|
ASSERT_TRUE(Overlaps("375", "400"));
|
|
|
|
ASSERT_TRUE(Overlaps("450", "450"));
|
|
|
|
ASSERT_TRUE(Overlaps("450", "500"));
|
|
|
|
ASSERT_TRUE(Overlaps("450", "700"));
|
|
|
|
ASSERT_TRUE(Overlaps("600", "700"));
|
|
|
|
}
|
|
|
|
|
2018-06-28 21:16:10 +02:00
|
|
|
class ManifestWriterTest : public testing::Test {
|
|
|
|
public:
|
|
|
|
ManifestWriterTest()
|
|
|
|
: env_(Env::Default()),
|
2018-07-14 02:18:39 +02:00
|
|
|
dbname_(test::PerThreadDBPath("version_set_test")),
|
2018-06-28 21:16:10 +02:00
|
|
|
db_options_(),
|
|
|
|
mutable_cf_options_(cf_options_),
|
|
|
|
table_cache_(NewLRUCache(50000, 16)),
|
|
|
|
write_buffer_manager_(db_options_.db_write_buffer_size),
|
|
|
|
versions_(new VersionSet(dbname_, &db_options_, env_options_,
|
|
|
|
table_cache_.get(), &write_buffer_manager_,
|
|
|
|
&write_controller_)),
|
|
|
|
shutting_down_(false),
|
|
|
|
mock_table_factory_(std::make_shared<mock::MockTableFactory>()) {
|
|
|
|
EXPECT_OK(env_->CreateDirIfMissing(dbname_));
|
|
|
|
db_options_.db_paths.emplace_back(dbname_,
|
|
|
|
std::numeric_limits<uint64_t>::max());
|
|
|
|
}
|
|
|
|
|
|
|
|
// Create DB with 3 column families.
|
|
|
|
void NewDB() {
|
|
|
|
VersionEdit new_db;
|
|
|
|
new_db.SetLogNumber(0);
|
|
|
|
new_db.SetNextFile(2);
|
|
|
|
new_db.SetLastSequence(0);
|
|
|
|
|
|
|
|
const std::vector<std::string> cf_names = {kDefaultColumnFamilyName,
|
|
|
|
"alice", "bob"};
|
|
|
|
const int kInitialNumOfCfs = static_cast<int>(cf_names.size());
|
|
|
|
autovector<VersionEdit> new_cfs;
|
|
|
|
uint64_t last_seq = 1;
|
|
|
|
uint32_t cf_id = 1;
|
|
|
|
for (int i = 1; i != kInitialNumOfCfs; ++i) {
|
|
|
|
VersionEdit new_cf;
|
|
|
|
new_cf.AddColumnFamily(cf_names[i]);
|
|
|
|
new_cf.SetColumnFamily(cf_id++);
|
|
|
|
new_cf.SetLogNumber(0);
|
|
|
|
new_cf.SetNextFile(2);
|
|
|
|
new_cf.SetLastSequence(last_seq++);
|
|
|
|
new_cfs.emplace_back(new_cf);
|
|
|
|
}
|
|
|
|
|
|
|
|
const std::string manifest = DescriptorFileName(dbname_, 1);
|
|
|
|
unique_ptr<WritableFile> file;
|
|
|
|
Status s = env_->NewWritableFile(
|
|
|
|
manifest, &file, env_->OptimizeForManifestWrite(env_options_));
|
|
|
|
ASSERT_OK(s);
|
|
|
|
unique_ptr<WritableFileWriter> file_writer(
|
|
|
|
new WritableFileWriter(std::move(file), env_options_));
|
|
|
|
{
|
|
|
|
log::Writer log(std::move(file_writer), 0, false);
|
|
|
|
std::string record;
|
|
|
|
new_db.EncodeTo(&record);
|
|
|
|
s = log.AddRecord(record);
|
|
|
|
for (const auto& e : new_cfs) {
|
|
|
|
e.EncodeTo(&record);
|
|
|
|
s = log.AddRecord(record);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
ASSERT_OK(s);
|
|
|
|
// Make "CURRENT" file point to the new manifest file.
|
|
|
|
s = SetCurrentFile(env_, dbname_, 1, nullptr);
|
|
|
|
|
|
|
|
std::vector<ColumnFamilyDescriptor> column_families;
|
|
|
|
cf_options_.table_factory = mock_table_factory_;
|
|
|
|
for (const auto& cf_name : cf_names) {
|
|
|
|
column_families.emplace_back(cf_name, cf_options_);
|
|
|
|
}
|
|
|
|
|
|
|
|
EXPECT_OK(versions_->Recover(column_families, false));
|
|
|
|
EXPECT_EQ(kInitialNumOfCfs,
|
|
|
|
versions_->GetColumnFamilySet()->NumberOfColumnFamilies());
|
|
|
|
for (auto cfd : *versions_->GetColumnFamilySet()) {
|
|
|
|
cfds_.emplace_back(cfd);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Env* env_;
|
|
|
|
const std::string dbname_;
|
|
|
|
EnvOptions env_options_;
|
|
|
|
ImmutableDBOptions db_options_;
|
|
|
|
ColumnFamilyOptions cf_options_;
|
|
|
|
MutableCFOptions mutable_cf_options_;
|
|
|
|
std::shared_ptr<Cache> table_cache_;
|
|
|
|
WriteController write_controller_;
|
|
|
|
WriteBufferManager write_buffer_manager_;
|
|
|
|
std::unique_ptr<VersionSet> versions_;
|
|
|
|
InstrumentedMutex mutex_;
|
|
|
|
std::atomic<bool> shutting_down_;
|
|
|
|
std::shared_ptr<mock::MockTableFactory> mock_table_factory_;
|
|
|
|
std::vector<ColumnFamilyData*> cfds_;
|
|
|
|
};
|
|
|
|
|
|
|
|
TEST_F(ManifestWriterTest, SameColumnFamilyGroupCommit) {
|
|
|
|
NewDB();
|
|
|
|
const int kGroupSize = 5;
|
|
|
|
std::vector<VersionEdit> edits(kGroupSize);
|
|
|
|
std::vector<ColumnFamilyData*> cfds(kGroupSize, cfds_[0]);
|
|
|
|
std::vector<MutableCFOptions> all_mutable_cf_options(kGroupSize,
|
|
|
|
mutable_cf_options_);
|
|
|
|
std::vector<autovector<VersionEdit*>> edit_lists(kGroupSize);
|
|
|
|
for (int i = 0; i != kGroupSize; ++i) {
|
|
|
|
edit_lists[i].emplace_back(&edits[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
int count = 0;
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"VersionSet::ProcessManifestWrites:SameColumnFamily", [&](void* arg) {
|
|
|
|
uint32_t* cf_id = reinterpret_cast<uint32_t*>(arg);
|
|
|
|
EXPECT_EQ(0, *cf_id);
|
|
|
|
++count;
|
|
|
|
});
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
mutex_.Lock();
|
|
|
|
Status s =
|
|
|
|
versions_->LogAndApply(cfds, all_mutable_cf_options, edit_lists, &mutex_);
|
|
|
|
mutex_.Unlock();
|
|
|
|
EXPECT_OK(s);
|
|
|
|
EXPECT_EQ(kGroupSize - 1, count);
|
|
|
|
}
|
2013-10-04 06:49:15 +02:00
|
|
|
} // namespace rocksdb
|
2011-06-22 20:45:39 +02:00
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
2015-03-17 22:08:00 +01:00
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
return RUN_ALL_TESTS();
|
2011-06-22 20:45:39 +02:00
|
|
|
}
|