1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
|
// Copyright (c) 2013, 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
#ifndef OS_WIN
#include <unistd.h>
#endif // ! OS_WIN
#include <atomic>
#include <list>
#include <memory>
#include <set>
#include <sstream>
#include <stdexcept>
#include <string>
#include <thread>
#include "memory/arena.h"
#include "memtable/skiplist.h"
#include "monitoring/histogram.h"
#include "port/port.h"
#include "rocksdb/cache.h"
#include "rocksdb/comparator.h"
#include "rocksdb/persistent_cache.h"
#include "rocksdb/system_clock.h"
#include "util/coding.h"
#include "util/crc32c.h"
#include "util/mutexlock.h"
#include "utilities/persistent_cache/block_cache_tier_file.h"
#include "utilities/persistent_cache/block_cache_tier_metadata.h"
#include "utilities/persistent_cache/persistent_cache_util.h"
namespace ROCKSDB_NAMESPACE {
//
// Block cache tier implementation
//
class BlockCacheTier : public PersistentCacheTier {
public:
explicit BlockCacheTier(const PersistentCacheConfig& opt)
: opt_(opt),
insert_ops_(static_cast<size_t>(opt_.max_write_pipeline_backlog_size)),
buffer_allocator_(opt.write_buffer_size, opt.write_buffer_count()),
writer_(this, opt_.writer_qdepth,
static_cast<size_t>(opt_.writer_dispatch_size)) {
Info(opt_.log, "Initializing allocator. size=%d B count=%" ROCKSDB_PRIszt,
opt_.write_buffer_size, opt_.write_buffer_count());
}
virtual ~BlockCacheTier() {
// Close is re-entrant so we can call close even if it is already closed
Close().PermitUncheckedError();
assert(!insert_th_.joinable());
}
Status Insert(const Slice& key, const char* data, const size_t size) override;
Status Lookup(const Slice& key, std::unique_ptr<char[]>* data,
size_t* size) override;
Status Open() override;
Status Close() override;
bool Erase(const Slice& key) override;
bool Reserve(const size_t size) override;
bool IsCompressed() override { return opt_.is_compressed; }
std::string GetPrintableOptions() const override { return opt_.ToString(); }
PersistentCache::StatsType Stats() override;
void TEST_Flush() override {
while (insert_ops_.Size()) {
/* sleep override */
SystemClock::Default()->SleepForMicroseconds(1000000);
}
}
private:
// Percentage of cache to be evicted when the cache is full
static const size_t kEvictPct = 10;
// Max attempts to insert key, value to cache in pipelined mode
static const size_t kMaxRetry = 3;
// Pipelined operation
struct InsertOp {
explicit InsertOp(const bool signal) : signal_(signal) {}
explicit InsertOp(std::string&& key, const std::string& data)
: key_(std::move(key)), data_(data) {}
~InsertOp() {}
InsertOp() = delete;
InsertOp(InsertOp&& /*rhs*/) = default;
InsertOp& operator=(InsertOp&& rhs) = default;
// used for estimating size by bounded queue
size_t Size() { return data_.size() + key_.size(); }
std::string key_;
std::string data_;
bool signal_ = false; // signal to request processing thread to exit
};
// entry point for insert thread
void InsertMain();
// insert implementation
Status InsertImpl(const Slice& key, const Slice& data);
// Create a new cache file
Status NewCacheFile();
// Get cache directory path
std::string GetCachePath() const { return opt_.path + "/cache"; }
// Cleanup folder
Status CleanupCacheFolder(const std::string& folder);
// Statistics
struct Statistics {
HistogramImpl bytes_pipelined_;
HistogramImpl bytes_written_;
HistogramImpl bytes_read_;
HistogramImpl read_hit_latency_;
HistogramImpl read_miss_latency_;
HistogramImpl write_latency_;
std::atomic<uint64_t> cache_hits_{0};
std::atomic<uint64_t> cache_misses_{0};
std::atomic<uint64_t> cache_errors_{0};
std::atomic<uint64_t> insert_dropped_{0};
double CacheHitPct() const {
const auto lookups = cache_hits_ + cache_misses_;
return lookups ? 100 * cache_hits_ / static_cast<double>(lookups) : 0.0;
}
double CacheMissPct() const {
const auto lookups = cache_hits_ + cache_misses_;
return lookups ? 100 * cache_misses_ / static_cast<double>(lookups) : 0.0;
}
};
port::RWMutex lock_; // Synchronization
const PersistentCacheConfig opt_; // BlockCache options
BoundedQueue<InsertOp> insert_ops_; // Ops waiting for insert
ROCKSDB_NAMESPACE::port::Thread insert_th_; // Insert thread
uint32_t writer_cache_id_ = 0; // Current cache file identifier
WriteableCacheFile* cache_file_ = nullptr; // Current cache file reference
CacheWriteBufferAllocator buffer_allocator_; // Buffer provider
ThreadedWriter writer_; // Writer threads
BlockCacheTierMetadata metadata_; // Cache meta data manager
std::atomic<uint64_t> size_{0}; // Size of the cache
Statistics stats_; // Statistics
};
} // namespace ROCKSDB_NAMESPACE
|
