rocksdb/db/version_edit.cc
Dhruba Borthakur 47c4191fe8 Reduce write amplification by merging files in L0 back into L0
Summary:
There is a new option called hybrid_mode which, when switched on,
causes HBase style compactions.  Files from L0 are
compacted back into L0. This meat of this compaction algorithm
is in PickCompactionHybrid().

All files reside in L0. That means all files have overlapping
keys. Each file has a time-bound, i.e. each file contains a
range of keys that were inserted around the same time. The
start-seqno and the end-seqno refers to the timeframe when
these keys were inserted.  Files that have contiguous seqno
are compacted together into a larger file. All files are
ordered from most recent to the oldest.

The current compaction algorithm starts to look for
candidate files starting from the most recent file. It continues to
add more files to the same compaction run as long as the
sum of the files chosen till now is smaller than the next
candidate file size. This logic needs to be debated
and validated.

The above logic should reduce write amplification to a
large extent... will publish numbers shortly.

Test Plan: dbstress runs for 6 hours with no data corruption (tested so far).

Differential Revision: https://reviews.facebook.net/D11289
2013-06-30 20:07:04 -07:00

297 lines
7.8 KiB
C++

// 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_edit.h"
#include "db/version_set.h"
#include "util/coding.h"
namespace leveldb {
// Tag numbers for serialized VersionEdit. These numbers are written to
// disk and should not be changed.
enum Tag {
kComparator = 1,
kLogNumber = 2,
kNextFileNumber = 3,
kLastSequence = 4,
kCompactPointer = 5,
kDeletedFile = 6,
kNewFile = 7,
// 8 was used for large value refs
kPrevLogNumber = 9,
// these are new formats divergent from open source leveldb
kNewFile2 = 100 // store smallest & largest seqno
};
void VersionEdit::Clear() {
comparator_.clear();
log_number_ = 0;
prev_log_number_ = 0;
last_sequence_ = 0;
next_file_number_ = 0;
has_comparator_ = false;
has_log_number_ = false;
has_prev_log_number_ = false;
has_next_file_number_ = false;
has_last_sequence_ = false;
deleted_files_.clear();
new_files_.clear();
}
void VersionEdit::EncodeTo(std::string* dst) const {
if (has_comparator_) {
PutVarint32(dst, kComparator);
PutLengthPrefixedSlice(dst, comparator_);
}
if (has_log_number_) {
PutVarint32(dst, kLogNumber);
PutVarint64(dst, log_number_);
}
if (has_prev_log_number_) {
PutVarint32(dst, kPrevLogNumber);
PutVarint64(dst, prev_log_number_);
}
if (has_next_file_number_) {
PutVarint32(dst, kNextFileNumber);
PutVarint64(dst, next_file_number_);
}
if (has_last_sequence_) {
PutVarint32(dst, kLastSequence);
PutVarint64(dst, last_sequence_);
}
for (size_t i = 0; i < compact_pointers_.size(); i++) {
PutVarint32(dst, kCompactPointer);
PutVarint32(dst, compact_pointers_[i].first); // level
PutLengthPrefixedSlice(dst, compact_pointers_[i].second.Encode());
}
for (DeletedFileSet::const_iterator iter = deleted_files_.begin();
iter != deleted_files_.end();
++iter) {
PutVarint32(dst, kDeletedFile);
PutVarint32(dst, iter->first); // level
PutVarint64(dst, iter->second); // file number
}
for (size_t i = 0; i < new_files_.size(); i++) {
const FileMetaData& f = new_files_[i].second;
PutVarint32(dst, kNewFile2);
PutVarint32(dst, new_files_[i].first); // level
PutVarint64(dst, f.number);
PutVarint64(dst, f.file_size);
PutLengthPrefixedSlice(dst, f.smallest.Encode());
PutLengthPrefixedSlice(dst, f.largest.Encode());
PutVarint64(dst, f.smallest_seqno);
PutVarint64(dst, f.largest_seqno);
}
}
static bool GetInternalKey(Slice* input, InternalKey* dst) {
Slice str;
if (GetLengthPrefixedSlice(input, &str)) {
dst->DecodeFrom(str);
return true;
} else {
return false;
}
}
bool VersionEdit::GetLevel(Slice* input, int* level, const char** msg) {
uint32_t v;
if (GetVarint32(input, &v) &&
(int)v < number_levels_) {
*level = v;
return true;
} else {
if ((int)v >= number_levels_) {
*msg = "db already has more levels than options.num_levels";
}
return false;
}
}
Status VersionEdit::DecodeFrom(const Slice& src) {
Clear();
Slice input = src;
const char* msg = nullptr;
uint32_t tag;
// Temporary storage for parsing
int level;
uint64_t number;
FileMetaData f;
Slice str;
InternalKey key;
while (msg == nullptr && GetVarint32(&input, &tag)) {
switch (tag) {
case kComparator:
if (GetLengthPrefixedSlice(&input, &str)) {
comparator_ = str.ToString();
has_comparator_ = true;
} else {
msg = "comparator name";
}
break;
case kLogNumber:
if (GetVarint64(&input, &log_number_)) {
has_log_number_ = true;
} else {
msg = "log number";
}
break;
case kPrevLogNumber:
if (GetVarint64(&input, &prev_log_number_)) {
has_prev_log_number_ = true;
} else {
msg = "previous log number";
}
break;
case kNextFileNumber:
if (GetVarint64(&input, &next_file_number_)) {
has_next_file_number_ = true;
} else {
msg = "next file number";
}
break;
case kLastSequence:
if (GetVarint64(&input, &last_sequence_)) {
has_last_sequence_ = true;
} else {
msg = "last sequence number";
}
break;
case kCompactPointer:
if (GetLevel(&input, &level, &msg) &&
GetInternalKey(&input, &key)) {
compact_pointers_.push_back(std::make_pair(level, key));
} else {
if (!msg) {
msg = "compaction pointer";
}
}
break;
case kDeletedFile:
if (GetLevel(&input, &level, &msg) &&
GetVarint64(&input, &number)) {
deleted_files_.insert(std::make_pair(level, number));
} else {
if (!msg) {
msg = "deleted file";
}
}
break;
case kNewFile:
if (GetLevel(&input, &level, &msg) &&
GetVarint64(&input, &f.number) &&
GetVarint64(&input, &f.file_size) &&
GetInternalKey(&input, &f.smallest) &&
GetInternalKey(&input, &f.largest)) {
new_files_.push_back(std::make_pair(level, f));
} else {
if (!msg) {
msg = "new-file entry";
}
}
break;
case kNewFile2:
if (GetLevel(&input, &level, &msg) &&
GetVarint64(&input, &f.number) &&
GetVarint64(&input, &f.file_size) &&
GetInternalKey(&input, &f.smallest) &&
GetInternalKey(&input, &f.largest) &&
GetVarint64(&input, &f.smallest_seqno) &&
GetVarint64(&input, &f.largest_seqno) ) {
new_files_.push_back(std::make_pair(level, f));
} else {
if (!msg) {
msg = "new-file2 entry";
}
}
break;
default:
msg = "unknown tag";
break;
}
}
if (msg == nullptr && !input.empty()) {
msg = "invalid tag";
}
Status result;
if (msg != nullptr) {
result = Status::Corruption("VersionEdit", msg);
}
return result;
}
std::string VersionEdit::DebugString() const {
std::string r;
r.append("VersionEdit {");
if (has_comparator_) {
r.append("\n Comparator: ");
r.append(comparator_);
}
if (has_log_number_) {
r.append("\n LogNumber: ");
AppendNumberTo(&r, log_number_);
}
if (has_prev_log_number_) {
r.append("\n PrevLogNumber: ");
AppendNumberTo(&r, prev_log_number_);
}
if (has_next_file_number_) {
r.append("\n NextFile: ");
AppendNumberTo(&r, next_file_number_);
}
if (has_last_sequence_) {
r.append("\n LastSeq: ");
AppendNumberTo(&r, last_sequence_);
}
for (size_t i = 0; i < compact_pointers_.size(); i++) {
r.append("\n CompactPointer: ");
AppendNumberTo(&r, compact_pointers_[i].first);
r.append(" ");
r.append(compact_pointers_[i].second.DebugString());
}
for (DeletedFileSet::const_iterator iter = deleted_files_.begin();
iter != deleted_files_.end();
++iter) {
r.append("\n DeleteFile: ");
AppendNumberTo(&r, iter->first);
r.append(" ");
AppendNumberTo(&r, iter->second);
}
for (size_t i = 0; i < new_files_.size(); i++) {
const FileMetaData& f = new_files_[i].second;
r.append("\n AddFile: ");
AppendNumberTo(&r, new_files_[i].first);
r.append(" ");
AppendNumberTo(&r, f.number);
r.append(" ");
AppendNumberTo(&r, f.file_size);
r.append(" ");
r.append(f.smallest.DebugString());
r.append(" .. ");
r.append(f.largest.DebugString());
}
r.append("\n}\n");
return r;
}
} // namespace leveldb