rocksdb/trace_replay/block_cache_tracer.h

212 lines
7.8 KiB
C
Raw Normal View History

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// 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).
#pragma once
#include <atomic>
#include "monitoring/instrumented_mutex.h"
#include "rocksdb/env.h"
#include "rocksdb/options.h"
#include "rocksdb/trace_reader_writer.h"
#include "table/table_reader_caller.h"
#include "trace_replay/trace_replay.h"
namespace rocksdb {
Create a BlockCacheLookupContext to enable fine-grained block cache tracing. (#5421) Summary: BlockCacheLookupContext only contains the caller for now. We will trace block accesses at five places: 1. BlockBasedTable::GetFilter. 2. BlockBasedTable::GetUncompressedDict. 3. BlockBasedTable::MaybeReadAndLoadToCache. (To trace access on data, index, and range deletion block.) 4. BlockBasedTable::Get. (To trace the referenced key and whether the referenced key exists in a fetched data block.) 5. BlockBasedTable::MultiGet. (To trace the referenced key and whether the referenced key exists in a fetched data block.) We create the context at: 1. BlockBasedTable::Get. (kUserGet) 2. BlockBasedTable::MultiGet. (kUserMGet) 3. BlockBasedTable::NewIterator. (either kUserIterator, kCompaction, or external SST ingestion calls this function.) 4. BlockBasedTable::Open. (kPrefetch) 5. Index/Filter::CacheDependencies. (kPrefetch) 6. BlockBasedTable::ApproximateOffsetOf. (kCompaction or kUserApproximateSize). I loaded 1 million key-value pairs into the database and ran the readrandom benchmark with a single thread. I gave the block cache 10 GB to make sure all reads hit the block cache after warmup. The throughput is comparable. Throughput of this PR: 231334 ops/s. Throughput of the master branch: 238428 ops/s. Experiment setup: RocksDB: version 6.2 Date: Mon Jun 10 10:42:51 2019 CPU: 24 * Intel Core Processor (Skylake) CPUCache: 16384 KB Keys: 20 bytes each Values: 100 bytes each (100 bytes after compression) Entries: 1000000 Prefix: 20 bytes Keys per prefix: 0 RawSize: 114.4 MB (estimated) FileSize: 114.4 MB (estimated) Write rate: 0 bytes/second Read rate: 0 ops/second Compression: NoCompression Compression sampling rate: 0 Memtablerep: skip_list Perf Level: 1 Load command: ./db_bench --benchmarks="fillseq" --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --statistics --cache_index_and_filter_blocks --cache_size=10737418240 --disable_auto_compactions=1 --disable_wal=1 --compression_type=none --min_level_to_compress=-1 --compression_ratio=1 --num=1000000 Run command: ./db_bench --benchmarks="readrandom,stats" --use_existing_db --threads=1 --duration=120 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --statistics --cache_index_and_filter_blocks --cache_size=10737418240 --disable_auto_compactions=1 --disable_wal=1 --compression_type=none --min_level_to_compress=-1 --compression_ratio=1 --num=1000000 --duration=120 TODOs: 1. Create a caller for external SST file ingestion and differentiate the callers for iterator. 2. Integrate tracer to trace block cache accesses. Pull Request resolved: https://github.com/facebook/rocksdb/pull/5421 Differential Revision: D15704258 Pulled By: HaoyuHuang fbshipit-source-id: 4aa8a55f8cb1576ffb367bfa3186a91d8f06d93a
2019-06-11 00:30:05 +02:00
// Lookup context for tracing block cache accesses.
// We trace block accesses at five places:
// 1. BlockBasedTable::GetFilter
// 2. BlockBasedTable::GetUncompressedDict.
// 3. BlockBasedTable::MaybeReadAndLoadToCache. (To trace access on data, index,
// and range deletion block.)
// 4. BlockBasedTable::Get. (To trace the referenced key and whether the
// referenced key exists in a fetched data block.)
// 5. BlockBasedTable::MultiGet. (To trace the referenced key and whether the
// referenced key exists in a fetched data block.)
// The context is created at:
// 1. BlockBasedTable::Get. (kUserGet)
// 2. BlockBasedTable::MultiGet. (kUserMGet)
// 3. BlockBasedTable::NewIterator. (either kUserIterator, kCompaction, or
// external SST ingestion calls this function.)
// 4. BlockBasedTable::Open. (kPrefetch)
// 5. Index/Filter::CacheDependencies. (kPrefetch)
// 6. BlockBasedTable::ApproximateOffsetOf. (kCompaction or
// kUserApproximateSize).
struct BlockCacheLookupContext {
BlockCacheLookupContext(const TableReaderCaller& _caller) : caller(_caller) {}
const TableReaderCaller caller;
// These are populated when we perform lookup/insert on block cache. The block
// cache tracer uses these inforation when logging the block access at
// BlockBasedTable::GET and BlockBasedTable::MultiGet.
bool is_cache_hit = false;
bool no_insert = false;
TraceType block_type = TraceType::kTraceMax;
uint64_t block_size = 0;
std::string block_key;
uint64_t num_keys_in_block = 0;
void FillLookupContext(bool _is_cache_hit, bool _no_insert,
TraceType _block_type, uint64_t _block_size,
const std::string& _block_key,
uint64_t _num_keys_in_block) {
is_cache_hit = _is_cache_hit;
no_insert = _no_insert;
block_type = _block_type;
block_size = _block_size;
block_key = _block_key;
num_keys_in_block = _num_keys_in_block;
}
Create a BlockCacheLookupContext to enable fine-grained block cache tracing. (#5421) Summary: BlockCacheLookupContext only contains the caller for now. We will trace block accesses at five places: 1. BlockBasedTable::GetFilter. 2. BlockBasedTable::GetUncompressedDict. 3. BlockBasedTable::MaybeReadAndLoadToCache. (To trace access on data, index, and range deletion block.) 4. BlockBasedTable::Get. (To trace the referenced key and whether the referenced key exists in a fetched data block.) 5. BlockBasedTable::MultiGet. (To trace the referenced key and whether the referenced key exists in a fetched data block.) We create the context at: 1. BlockBasedTable::Get. (kUserGet) 2. BlockBasedTable::MultiGet. (kUserMGet) 3. BlockBasedTable::NewIterator. (either kUserIterator, kCompaction, or external SST ingestion calls this function.) 4. BlockBasedTable::Open. (kPrefetch) 5. Index/Filter::CacheDependencies. (kPrefetch) 6. BlockBasedTable::ApproximateOffsetOf. (kCompaction or kUserApproximateSize). I loaded 1 million key-value pairs into the database and ran the readrandom benchmark with a single thread. I gave the block cache 10 GB to make sure all reads hit the block cache after warmup. The throughput is comparable. Throughput of this PR: 231334 ops/s. Throughput of the master branch: 238428 ops/s. Experiment setup: RocksDB: version 6.2 Date: Mon Jun 10 10:42:51 2019 CPU: 24 * Intel Core Processor (Skylake) CPUCache: 16384 KB Keys: 20 bytes each Values: 100 bytes each (100 bytes after compression) Entries: 1000000 Prefix: 20 bytes Keys per prefix: 0 RawSize: 114.4 MB (estimated) FileSize: 114.4 MB (estimated) Write rate: 0 bytes/second Read rate: 0 ops/second Compression: NoCompression Compression sampling rate: 0 Memtablerep: skip_list Perf Level: 1 Load command: ./db_bench --benchmarks="fillseq" --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --statistics --cache_index_and_filter_blocks --cache_size=10737418240 --disable_auto_compactions=1 --disable_wal=1 --compression_type=none --min_level_to_compress=-1 --compression_ratio=1 --num=1000000 Run command: ./db_bench --benchmarks="readrandom,stats" --use_existing_db --threads=1 --duration=120 --key_size=20 --prefix_size=20 --keys_per_prefix=0 --value_size=100 --statistics --cache_index_and_filter_blocks --cache_size=10737418240 --disable_auto_compactions=1 --disable_wal=1 --compression_type=none --min_level_to_compress=-1 --compression_ratio=1 --num=1000000 --duration=120 TODOs: 1. Create a caller for external SST file ingestion and differentiate the callers for iterator. 2. Integrate tracer to trace block cache accesses. Pull Request resolved: https://github.com/facebook/rocksdb/pull/5421 Differential Revision: D15704258 Pulled By: HaoyuHuang fbshipit-source-id: 4aa8a55f8cb1576ffb367bfa3186a91d8f06d93a
2019-06-11 00:30:05 +02:00
};
enum Boolean : char { kTrue = 1, kFalse = 0 };
struct BlockCacheTraceRecord {
// Required fields for all accesses.
uint64_t access_timestamp = 0;
std::string block_key;
TraceType block_type = TraceType::kTraceMax;
uint64_t block_size = 0;
uint64_t cf_id = 0;
std::string cf_name;
uint32_t level = 0;
uint64_t sst_fd_number = 0;
TableReaderCaller caller = TableReaderCaller::kMaxBlockCacheLookupCaller;
Boolean is_cache_hit = Boolean::kFalse;
Boolean no_insert = Boolean::kFalse;
// Required fields for data block and user Get/Multi-Get only.
std::string referenced_key;
uint64_t referenced_data_size = 0;
uint64_t num_keys_in_block = 0;
Boolean referenced_key_exist_in_block = Boolean::kFalse;
BlockCacheTraceRecord() {}
BlockCacheTraceRecord(uint64_t _access_timestamp, std::string _block_key,
TraceType _block_type, uint64_t _block_size,
uint64_t _cf_id, std::string _cf_name, uint32_t _level,
uint64_t _sst_fd_number, TableReaderCaller _caller,
bool _is_cache_hit, bool _no_insert,
std::string _referenced_key = "",
uint64_t _referenced_data_size = 0,
uint64_t _num_keys_in_block = 0,
bool _referenced_key_exist_in_block = false)
: access_timestamp(_access_timestamp),
block_key(_block_key),
block_type(_block_type),
block_size(_block_size),
cf_id(_cf_id),
cf_name(_cf_name),
level(_level),
sst_fd_number(_sst_fd_number),
caller(_caller),
is_cache_hit(_is_cache_hit ? Boolean::kTrue : Boolean::kFalse),
no_insert(_no_insert ? Boolean::kTrue : Boolean::kFalse),
referenced_key(_referenced_key),
referenced_data_size(_referenced_data_size),
num_keys_in_block(_num_keys_in_block),
referenced_key_exist_in_block(
_referenced_key_exist_in_block ? Boolean::kTrue : Boolean::kFalse) {
}
};
struct BlockCacheTraceHeader {
uint64_t start_time;
uint32_t rocksdb_major_version;
uint32_t rocksdb_minor_version;
};
class BlockCacheTraceHelper {
public:
static bool ShouldTraceReferencedKey(TraceType block_type,
TableReaderCaller caller);
static const std::string kUnknownColumnFamilyName;
};
// BlockCacheTraceWriter captures all RocksDB block cache accesses using a
// user-provided TraceWriter. Every RocksDB operation is written as a single
// trace. Each trace will have a timestamp and type, followed by the trace
// payload.
class BlockCacheTraceWriter {
public:
BlockCacheTraceWriter(Env* env, const TraceOptions& trace_options,
std::unique_ptr<TraceWriter>&& trace_writer);
~BlockCacheTraceWriter() = default;
// No copy and move.
BlockCacheTraceWriter(const BlockCacheTraceWriter&) = delete;
BlockCacheTraceWriter& operator=(const BlockCacheTraceWriter&) = delete;
BlockCacheTraceWriter(BlockCacheTraceWriter&&) = delete;
BlockCacheTraceWriter& operator=(BlockCacheTraceWriter&&) = delete;
// Pass Slice references to avoid copy.
Status WriteBlockAccess(const BlockCacheTraceRecord& record,
const Slice& block_key, const Slice& cf_name,
const Slice& referenced_key);
// Write a trace header at the beginning, typically on initiating a trace,
// with some metadata like a magic number and RocksDB version.
Status WriteHeader();
private:
Env* env_;
TraceOptions trace_options_;
std::unique_ptr<TraceWriter> trace_writer_;
};
// BlockCacheTraceReader helps read the trace file generated by
// BlockCacheTraceWriter using a user provided TraceReader.
class BlockCacheTraceReader {
public:
BlockCacheTraceReader(std::unique_ptr<TraceReader>&& reader);
~BlockCacheTraceReader() = default;
// No copy and move.
BlockCacheTraceReader(const BlockCacheTraceReader&) = delete;
BlockCacheTraceReader& operator=(const BlockCacheTraceReader&) = delete;
BlockCacheTraceReader(BlockCacheTraceReader&&) = delete;
BlockCacheTraceReader& operator=(BlockCacheTraceReader&&) = delete;
Status ReadHeader(BlockCacheTraceHeader* header);
Status ReadAccess(BlockCacheTraceRecord* record);
private:
std::unique_ptr<TraceReader> trace_reader_;
};
// A block cache tracer. It downsamples the accesses according to
// trace_options and uses BlockCacheTraceWriter to write the access record to
// the trace file.
class BlockCacheTracer {
public:
BlockCacheTracer();
~BlockCacheTracer();
// No copy and move.
BlockCacheTracer(const BlockCacheTracer&) = delete;
BlockCacheTracer& operator=(const BlockCacheTracer&) = delete;
BlockCacheTracer(BlockCacheTracer&&) = delete;
BlockCacheTracer& operator=(BlockCacheTracer&&) = delete;
// Start writing block cache accesses to the trace_writer.
Status StartTrace(Env* env, const TraceOptions& trace_options,
std::unique_ptr<TraceWriter>&& trace_writer);
// Stop writing block cache accesses to the trace_writer.
void EndTrace();
Status WriteBlockAccess(const BlockCacheTraceRecord& record,
const Slice& block_key, const Slice& cf_name,
const Slice& referenced_key);
private:
TraceOptions trace_options_;
// A mutex protects the writer_.
InstrumentedMutex trace_writer_mutex_;
std::atomic<BlockCacheTraceWriter*> writer_;
};
} // namespace rocksdb