rocksdb/db/file_indexer.cc
Feng Zhu c11d604ab3 store file_indexer info in sequential memory
Summary:
  use arena to allocate space for next_level_index_ and level_rb_
  Thus increasing data locality and make Version::Get faster.

Benchmark detail
Base version: commit d2a727c182

command used:
./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=2097152 --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=fillseq, readrandom,readrandom,readrandom --use_existing_db=0 --num=52428800 --threads=1

Result:
cpu running percentage:
Version::Get, improved from 7.98% to 7.42%
FileIndexer::GetNextLevelIndex, improved from 1.18% to 0.68%.

Test Plan:
  make all check

Reviewers: ljin, haobo, yhchiang, sdong

Reviewed By: sdong

Subscribers: dhruba, igor

Differential Revision: https://reviews.facebook.net/D19845
2014-07-16 11:21:30 -07:00

218 lines
7.3 KiB
C++

// Copyright (c) 2013, 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 "db/file_indexer.h"
#include <algorithm>
#include "rocksdb/comparator.h"
#include "db/version_edit.h"
namespace rocksdb {
FileIndexer::FileIndexer(const Comparator* ucmp)
: num_levels_(0),
ucmp_(ucmp),
level_rb_(nullptr) {
}
uint32_t FileIndexer::NumLevelIndex() {
return next_level_index_.size();
}
uint32_t FileIndexer::LevelIndexSize(uint32_t level) {
return next_level_index_[level].num_index;
}
void FileIndexer::GetNextLevelIndex(
const uint32_t level, const uint32_t file_index, const int cmp_smallest,
const int cmp_largest, int32_t* left_bound, int32_t* right_bound) {
assert(level > 0);
// Last level, no hint
if (level == num_levels_ - 1) {
*left_bound = 0;
*right_bound = -1;
return;
}
assert(level < num_levels_ - 1);
assert(static_cast<int32_t>(file_index) <= level_rb_[level]);
const IndexUnit* index_units = next_level_index_[level].index_units;
const auto& index = index_units[file_index];
if (cmp_smallest < 0) {
*left_bound = (level > 0 && file_index > 0) ?
index_units[file_index - 1].largest_lb : 0;
*right_bound = index.smallest_rb;
} else if (cmp_smallest == 0) {
*left_bound = index.smallest_lb;
*right_bound = index.smallest_rb;
} else if (cmp_smallest > 0 && cmp_largest < 0) {
*left_bound = index.smallest_lb;
*right_bound = index.largest_rb;
} else if (cmp_largest == 0) {
*left_bound = index.largest_lb;
*right_bound = index.largest_rb;
} else if (cmp_largest > 0) {
*left_bound = index.largest_lb;
*right_bound = level_rb_[level + 1];
} else {
assert(false);
}
assert(*left_bound >= 0);
assert(*left_bound <= *right_bound + 1);
assert(*right_bound <= level_rb_[level + 1]);
}
void FileIndexer::UpdateIndex(Arena* arena,
const uint32_t num_levels,
std::vector<FileMetaData*>* const files) {
if (files == nullptr) {
return;
}
if (num_levels == 0) { // uint_32 0-1 would cause bad behavior
num_levels_ = num_levels;
return;
}
assert(level_rb_ == nullptr); // level_rb_ should be init here
num_levels_ = num_levels;
next_level_index_.resize(num_levels);
char* mem = arena->AllocateAligned(num_levels_ * sizeof(int32_t));
level_rb_ = new (mem)int32_t[num_levels_];
for (size_t i = 0; i < num_levels_; i++) {
level_rb_[i] = -1;
}
// L1 - Ln-1
for (uint32_t level = 1; level < num_levels_ - 1; ++level) {
const auto& upper_files = files[level];
const int32_t upper_size = upper_files.size();
const auto& lower_files = files[level + 1];
level_rb_[level] = upper_files.size() - 1;
if (upper_size == 0) {
continue;
}
IndexLevel& index_level = next_level_index_[level];
index_level.num_index = upper_size;
char* mem = arena->AllocateAligned(upper_size * sizeof(IndexUnit));
index_level.index_units = new (mem)IndexUnit[upper_size];
CalculateLB(upper_files, lower_files, &index_level,
[this](const FileMetaData* a, const FileMetaData* b) -> int {
return ucmp_->Compare(a->smallest.user_key(), b->largest.user_key());
},
[](IndexUnit* index, int32_t f_idx) {
index->smallest_lb = f_idx;
});
CalculateLB(upper_files, lower_files, &index_level,
[this](const FileMetaData* a, const FileMetaData* b) -> int {
return ucmp_->Compare(a->largest.user_key(), b->largest.user_key());
},
[](IndexUnit* index, int32_t f_idx) {
index->largest_lb = f_idx;
});
CalculateRB(upper_files, lower_files, &index_level,
[this](const FileMetaData* a, const FileMetaData* b) -> int {
return ucmp_->Compare(a->smallest.user_key(), b->smallest.user_key());
},
[](IndexUnit* index, int32_t f_idx) {
index->smallest_rb = f_idx;
});
CalculateRB(upper_files, lower_files, &index_level,
[this](const FileMetaData* a, const FileMetaData* b) -> int {
return ucmp_->Compare(a->largest.user_key(), b->smallest.user_key());
},
[](IndexUnit* index, int32_t f_idx) {
index->largest_rb = f_idx;
});
}
level_rb_[num_levels_ - 1] = files[num_levels_ - 1].size() - 1;
}
void FileIndexer::CalculateLB(const std::vector<FileMetaData*>& upper_files,
const std::vector<FileMetaData*>& lower_files,
IndexLevel *index_level,
std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
std::function<void(IndexUnit*, int32_t)> set_index) {
const int32_t upper_size = upper_files.size();
const int32_t lower_size = lower_files.size();
int32_t upper_idx = 0;
int32_t lower_idx = 0;
IndexUnit* index = index_level->index_units;
while (upper_idx < upper_size && lower_idx < lower_size) {
int cmp = cmp_op(upper_files[upper_idx], lower_files[lower_idx]);
if (cmp == 0) {
set_index(&index[upper_idx], lower_idx);
++upper_idx;
++lower_idx;
} else if (cmp > 0) {
// Lower level's file (largest) is smaller, a key won't hit in that
// file. Move to next lower file
++lower_idx;
} else {
// Lower level's file becomes larger, update the index, and
// move to the next upper file
set_index(&index[upper_idx], lower_idx);
++upper_idx;
}
}
while (upper_idx < upper_size) {
// Lower files are exhausted, that means the remaining upper files are
// greater than any lower files. Set the index to be the lower level size.
set_index(&index[upper_idx], lower_size);
++upper_idx;
}
}
void FileIndexer::CalculateRB(const std::vector<FileMetaData*>& upper_files,
const std::vector<FileMetaData*>& lower_files,
IndexLevel *index_level,
std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
std::function<void(IndexUnit*, int32_t)> set_index) {
const int32_t upper_size = upper_files.size();
const int32_t lower_size = lower_files.size();
int32_t upper_idx = upper_size - 1;
int32_t lower_idx = lower_size - 1;
IndexUnit* index = index_level->index_units;
while (upper_idx >= 0 && lower_idx >= 0) {
int cmp = cmp_op(upper_files[upper_idx], lower_files[lower_idx]);
if (cmp == 0) {
set_index(&index[upper_idx], lower_idx);
--upper_idx;
--lower_idx;
} else if (cmp < 0) {
// Lower level's file (smallest) is larger, a key won't hit in that
// file. Move to next lower file.
--lower_idx;
} else {
// Lower level's file becomes smaller, update the index, and move to
// the next the upper file
set_index(&index[upper_idx], lower_idx);
--upper_idx;
}
}
while (upper_idx >= 0) {
// Lower files are exhausted, that means the remaining upper files are
// smaller than any lower files. Set it to -1.
set_index(&index[upper_idx], -1);
--upper_idx;
}
}
} // namespace rocksdb