summaryrefslogtreecommitdiffstats
path: root/src/rocksdb/include/rocksdb/cache.h
blob: 77ddf525d0b154ed7b41969135b3a10790d174c8 (plain)
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
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
// 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).
//
// 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.
//
// A Cache is an interface that maps keys to values.  It has internal
// synchronization and may be safely accessed concurrently from
// multiple threads.  It may automatically evict entries to make room
// for new entries.  Values have a specified charge against the cache
// capacity.  For example, a cache where the values are variable
// length strings, may use the length of the string as the charge for
// the string.
//
// A builtin cache implementation with a least-recently-used eviction
// policy is provided.  Clients may use their own implementations if
// they want something more sophisticated (like scan-resistance, a
// custom eviction policy, variable cache sizing, etc.)

#pragma once

#include <stdint.h>
#include <memory>
#include <string>
#include "rocksdb/memory_allocator.h"
#include "rocksdb/slice.h"
#include "rocksdb/statistics.h"
#include "rocksdb/status.h"

namespace ROCKSDB_NAMESPACE {

class Cache;

extern const bool kDefaultToAdaptiveMutex;

enum CacheMetadataChargePolicy {
  kDontChargeCacheMetadata,
  kFullChargeCacheMetadata
};
const CacheMetadataChargePolicy kDefaultCacheMetadataChargePolicy =
    kFullChargeCacheMetadata;

struct LRUCacheOptions {
  // Capacity of the cache.
  size_t capacity = 0;

  // Cache is sharded into 2^num_shard_bits shards,
  // by hash of key. Refer to NewLRUCache for further
  // information.
  int num_shard_bits = -1;

  // If strict_capacity_limit is set,
  // insert to the cache will fail when cache is full.
  bool strict_capacity_limit = false;

  // Percentage of cache reserved for high priority entries.
  // If greater than zero, the LRU list will be split into a high-pri
  // list and a low-pri list. High-pri entries will be insert to the
  // tail of high-pri list, while low-pri entries will be first inserted to
  // the low-pri list (the midpoint). This is refered to as
  // midpoint insertion strategy to make entries never get hit in cache
  // age out faster.
  //
  // See also
  // BlockBasedTableOptions::cache_index_and_filter_blocks_with_high_priority.
  double high_pri_pool_ratio = 0.5;

  // If non-nullptr will use this allocator instead of system allocator when
  // allocating memory for cache blocks. Call this method before you start using
  // the cache!
  //
  // Caveat: when the cache is used as block cache, the memory allocator is
  // ignored when dealing with compression libraries that allocate memory
  // internally (currently only XPRESS).
  std::shared_ptr<MemoryAllocator> memory_allocator;

  // Whether to use adaptive mutexes for cache shards. Note that adaptive
  // mutexes need to be supported by the platform in order for this to have any
  // effect. The default value is true if RocksDB is compiled with
  // -DROCKSDB_DEFAULT_TO_ADAPTIVE_MUTEX, false otherwise.
  bool use_adaptive_mutex = kDefaultToAdaptiveMutex;

  CacheMetadataChargePolicy metadata_charge_policy =
      kDefaultCacheMetadataChargePolicy;

  LRUCacheOptions() {}
  LRUCacheOptions(size_t _capacity, int _num_shard_bits,
                  bool _strict_capacity_limit, double _high_pri_pool_ratio,
                  std::shared_ptr<MemoryAllocator> _memory_allocator = nullptr,
                  bool _use_adaptive_mutex = kDefaultToAdaptiveMutex,
                  CacheMetadataChargePolicy _metadata_charge_policy =
                      kDefaultCacheMetadataChargePolicy)
      : capacity(_capacity),
        num_shard_bits(_num_shard_bits),
        strict_capacity_limit(_strict_capacity_limit),
        high_pri_pool_ratio(_high_pri_pool_ratio),
        memory_allocator(std::move(_memory_allocator)),
        use_adaptive_mutex(_use_adaptive_mutex),
        metadata_charge_policy(_metadata_charge_policy) {}
};

// Create a new cache with a fixed size capacity. The cache is sharded
// to 2^num_shard_bits shards, by hash of the key. The total capacity
// is divided and evenly assigned to each shard. If strict_capacity_limit
// is set, insert to the cache will fail when cache is full. User can also
// set percentage of the cache reserves for high priority entries via
// high_pri_pool_pct.
// num_shard_bits = -1 means it is automatically determined: every shard
// will be at least 512KB and number of shard bits will not exceed 6.
extern std::shared_ptr<Cache> NewLRUCache(
    size_t capacity, int num_shard_bits = -1,
    bool strict_capacity_limit = false, double high_pri_pool_ratio = 0.5,
    std::shared_ptr<MemoryAllocator> memory_allocator = nullptr,
    bool use_adaptive_mutex = kDefaultToAdaptiveMutex,
    CacheMetadataChargePolicy metadata_charge_policy =
        kDefaultCacheMetadataChargePolicy);

extern std::shared_ptr<Cache> NewLRUCache(const LRUCacheOptions& cache_opts);

// Similar to NewLRUCache, but create a cache based on CLOCK algorithm with
// better concurrent performance in some cases. See util/clock_cache.cc for
// more detail.
//
// Return nullptr if it is not supported.
extern std::shared_ptr<Cache> NewClockCache(
    size_t capacity, int num_shard_bits = -1,
    bool strict_capacity_limit = false,
    CacheMetadataChargePolicy metadata_charge_policy =
        kDefaultCacheMetadataChargePolicy);
class Cache {
 public:
  // Depending on implementation, cache entries with high priority could be less
  // likely to get evicted than low priority entries.
  enum class Priority { HIGH, LOW };

  Cache(std::shared_ptr<MemoryAllocator> allocator = nullptr)
      : memory_allocator_(std::move(allocator)) {}
  // No copying allowed
  Cache(const Cache&) = delete;
  Cache& operator=(const Cache&) = delete;

  // Destroys all existing entries by calling the "deleter"
  // function that was passed via the Insert() function.
  //
  // @See Insert
  virtual ~Cache() {}

  // Opaque handle to an entry stored in the cache.
  struct Handle {};

  // The type of the Cache
  virtual const char* Name() const = 0;

  // Insert a mapping from key->value into the cache and assign it
  // the specified charge against the total cache capacity.
  // If strict_capacity_limit is true and cache reaches its full capacity,
  // return Status::Incomplete.
  //
  // If handle is not nullptr, returns a handle that corresponds to the
  // mapping. The caller must call this->Release(handle) when the returned
  // mapping is no longer needed. In case of error caller is responsible to
  // cleanup the value (i.e. calling "deleter").
  //
  // If handle is nullptr, it is as if Release is called immediately after
  // insert. In case of error value will be cleanup.
  //
  // When the inserted entry is no longer needed, the key and
  // value will be passed to "deleter".
  virtual Status Insert(const Slice& key, void* value, size_t charge,
                        void (*deleter)(const Slice& key, void* value),
                        Handle** handle = nullptr,
                        Priority priority = Priority::LOW) = 0;

  // If the cache has no mapping for "key", returns nullptr.
  //
  // Else return a handle that corresponds to the mapping.  The caller
  // must call this->Release(handle) when the returned mapping is no
  // longer needed.
  // If stats is not nullptr, relative tickers could be used inside the
  // function.
  virtual Handle* Lookup(const Slice& key, Statistics* stats = nullptr) = 0;

  // Increments the reference count for the handle if it refers to an entry in
  // the cache. Returns true if refcount was incremented; otherwise, returns
  // false.
  // REQUIRES: handle must have been returned by a method on *this.
  virtual bool Ref(Handle* handle) = 0;

  /**
   * Release a mapping returned by a previous Lookup(). A released entry might
   * still  remain in cache in case it is later looked up by others. If
   * force_erase is set then it also erase it from the cache if there is no
   * other reference to  it. Erasing it should call the deleter function that
   * was provided when the
   * entry was inserted.
   *
   * Returns true if the entry was also erased.
   */
  // REQUIRES: handle must not have been released yet.
  // REQUIRES: handle must have been returned by a method on *this.
  virtual bool Release(Handle* handle, bool force_erase = false) = 0;

  // Return the value encapsulated in a handle returned by a
  // successful Lookup().
  // REQUIRES: handle must not have been released yet.
  // REQUIRES: handle must have been returned by a method on *this.
  virtual void* Value(Handle* handle) = 0;

  // If the cache contains entry for key, erase it.  Note that the
  // underlying entry will be kept around until all existing handles
  // to it have been released.
  virtual void Erase(const Slice& key) = 0;
  // Return a new numeric id.  May be used by multiple clients who are
  // sharding the same cache to partition the key space.  Typically the
  // client will allocate a new id at startup and prepend the id to
  // its cache keys.
  virtual uint64_t NewId() = 0;

  // sets the maximum configured capacity of the cache. When the new
  // capacity is less than the old capacity and the existing usage is
  // greater than new capacity, the implementation will do its best job to
  // purge the released entries from the cache in order to lower the usage
  virtual void SetCapacity(size_t capacity) = 0;

  // Set whether to return error on insertion when cache reaches its full
  // capacity.
  virtual void SetStrictCapacityLimit(bool strict_capacity_limit) = 0;

  // Get the flag whether to return error on insertion when cache reaches its
  // full capacity.
  virtual bool HasStrictCapacityLimit() const = 0;

  // returns the maximum configured capacity of the cache
  virtual size_t GetCapacity() const = 0;

  // returns the memory size for the entries residing in the cache.
  virtual size_t GetUsage() const = 0;

  // returns the memory size for a specific entry in the cache.
  virtual size_t GetUsage(Handle* handle) const = 0;

  // returns the memory size for the entries in use by the system
  virtual size_t GetPinnedUsage() const = 0;

  // returns the charge for the specific entry in the cache.
  virtual size_t GetCharge(Handle* handle) const = 0;

  // Call this on shutdown if you want to speed it up. Cache will disown
  // any underlying data and will not free it on delete. This call will leak
  // memory - call this only if you're shutting down the process.
  // Any attempts of using cache after this call will fail terribly.
  // Always delete the DB object before calling this method!
  virtual void DisownData(){
      // default implementation is noop
  }

  // Apply callback to all entries in the cache
  // If thread_safe is true, it will also lock the accesses. Otherwise, it will
  // access the cache without the lock held
  virtual void ApplyToAllCacheEntries(void (*callback)(void*, size_t),
                                      bool thread_safe) = 0;

  // Remove all entries.
  // Prerequisite: no entry is referenced.
  virtual void EraseUnRefEntries() = 0;

  virtual std::string GetPrintableOptions() const { return ""; }

  MemoryAllocator* memory_allocator() const { return memory_allocator_.get(); }

 private:
  std::shared_ptr<MemoryAllocator> memory_allocator_;
};

}  // namespace ROCKSDB_NAMESPACE