|
|
|
// 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.
|
|
|
|
|
|
|
|
#ifndef STORAGE_LEVELDB_TABLE_FORMAT_H_
|
|
|
|
#define STORAGE_LEVELDB_TABLE_FORMAT_H_
|
|
|
|
|
|
|
|
#include <string>
|
|
|
|
#include <stdint.h>
|
|
|
|
#include "rocksdb/slice.h"
|
|
|
|
#include "rocksdb/status.h"
|
|
|
|
#include "rocksdb/table_builder.h"
|
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
|
|
|
class Block;
|
|
|
|
class RandomAccessFile;
|
|
|
|
struct ReadOptions;
|
|
|
|
|
|
|
|
// BlockHandle is a pointer to the extent of a file that stores a data
|
|
|
|
// block or a meta block.
|
|
|
|
class BlockHandle {
|
|
|
|
public:
|
|
|
|
BlockHandle();
|
|
|
|
|
|
|
|
// The offset of the block in the file.
|
|
|
|
uint64_t offset() const { return offset_; }
|
|
|
|
void set_offset(uint64_t offset) { offset_ = offset; }
|
|
|
|
|
|
|
|
// The size of the stored block
|
|
|
|
uint64_t size() const { return size_; }
|
|
|
|
void set_size(uint64_t size) { size_ = size; }
|
|
|
|
|
|
|
|
void EncodeTo(std::string* dst) const;
|
|
|
|
Status DecodeFrom(Slice* input);
|
|
|
|
|
|
|
|
// Maximum encoding length of a BlockHandle
|
|
|
|
enum { kMaxEncodedLength = 10 + 10 };
|
|
|
|
|
|
|
|
private:
|
|
|
|
uint64_t offset_;
|
|
|
|
uint64_t size_;
|
|
|
|
};
|
|
|
|
|
|
|
|
// Footer encapsulates the fixed information stored at the tail
|
|
|
|
// end of every table file.
|
|
|
|
class Footer {
|
|
|
|
public:
|
|
|
|
Footer() { }
|
|
|
|
|
|
|
|
// The block handle for the metaindex block of the table
|
|
|
|
const BlockHandle& metaindex_handle() const { return metaindex_handle_; }
|
|
|
|
void set_metaindex_handle(const BlockHandle& h) { metaindex_handle_ = h; }
|
|
|
|
|
|
|
|
// The block handle for the index block of the table
|
|
|
|
const BlockHandle& index_handle() const {
|
|
|
|
return index_handle_;
|
|
|
|
}
|
|
|
|
void set_index_handle(const BlockHandle& h) {
|
|
|
|
index_handle_ = h;
|
|
|
|
}
|
|
|
|
|
|
|
|
void EncodeTo(std::string* dst) const;
|
|
|
|
Status DecodeFrom(Slice* input);
|
|
|
|
|
|
|
|
// Encoded length of a Footer. Note that the serialization of a
|
|
|
|
// Footer will always occupy exactly this many bytes. It consists
|
|
|
|
// of two block handles and a magic number.
|
|
|
|
enum {
|
|
|
|
kEncodedLength = 2*BlockHandle::kMaxEncodedLength + 8
|
|
|
|
};
|
|
|
|
|
|
|
|
private:
|
|
|
|
BlockHandle metaindex_handle_;
|
|
|
|
BlockHandle index_handle_;
|
|
|
|
};
|
|
|
|
|
|
|
|
// kTableMagicNumber was picked by running
|
|
|
|
// echo http://code.google.com/p/leveldb/ | sha1sum
|
|
|
|
// and taking the leading 64 bits.
|
|
|
|
static const uint64_t kTableMagicNumber = 0xdb4775248b80fb57ull;
|
|
|
|
|
|
|
|
// 1-byte type + 32-bit crc
|
|
|
|
static const size_t kBlockTrailerSize = 5;
|
|
|
|
|
|
|
|
struct BlockContents {
|
|
|
|
Slice data; // Actual contents of data
|
|
|
|
bool cachable; // True iff data can be cached
|
|
|
|
bool heap_allocated; // True iff caller should delete[] data.data()
|
|
|
|
};
|
|
|
|
|
|
|
|
// Read the block identified by "handle" from "file". On failure
|
|
|
|
// return non-OK. On success fill *result and return OK.
|
|
|
|
extern Status ReadBlockContents(RandomAccessFile* file,
|
|
|
|
const ReadOptions& options,
|
|
|
|
const BlockHandle& handle,
|
[RocksDB] Added nano second stopwatch and new perf counters to track block read cost
Summary: The pupose of this diff is to expose per user-call level precise timing of block read, so that we can answer questions like: a Get() costs me 100ms, is that somehow related to loading blocks from file system, or sth else? We will answer that with EXACTLY how many blocks have been read, how much time was spent on transfering the bytes from os, how much time was spent on checksum verification and how much time was spent on block decompression, just for that one Get. A nano second stopwatch was introduced to track time with higher precision. The cost/precision of the stopwatch is also measured in unit-test. On my dev box, retrieving one time instance costs about 30ns, on average. The deviation of timing results is good enough to track 100ns-1us level events. And the overhead could be safely ignored for 100us level events (10000 instances/s), for example, a viewstate thrift call.
Test Plan: perf_context_test, also testing with viewstate shadow traffic.
Reviewers: dhruba
Reviewed By: dhruba
CC: leveldb, xjin
Differential Revision: https://reviews.facebook.net/D12351
12 years ago
|
|
|
BlockContents* result,
|
|
|
|
Env* env);
|
|
|
|
|
|
|
|
// Implementation details follow. Clients should ignore,
|
|
|
|
|
|
|
|
inline BlockHandle::BlockHandle()
|
|
|
|
: offset_(~static_cast<uint64_t>(0)),
|
|
|
|
size_(~static_cast<uint64_t>(0)) {
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace rocksdb
|
|
|
|
|
|
|
|
#endif // STORAGE_LEVELDB_TABLE_FORMAT_H_
|