Expose usage info for the cache

Summary: This diff will help us to figure out the memory usage for the cache part.

Test Plan: added a new memory usage test for cache

Reviewers: haobo, sdong, dhruba

CC: leveldb

Differential Revision: https://reviews.facebook.net/D14559

include/rocksdb/cache.h

@@ -104,6 +104,9 @@ class Cache {
   // 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;
+
  private:
   void LRU_Remove(Handle* e);
   void LRU_Append(Handle* e);

util/cache.cc

@@ -156,6 +156,7 @@ class LRUCache {
   Cache::Handle* Lookup(const Slice& key, uint32_t hash);
   void Release(Cache::Handle* handle);
   void Erase(const Slice& key, uint32_t hash);
+  size_t GetUsage() const { return usage_.load(std::memory_order_relaxed); }
 
  private:
   void LRU_Remove(LRUHandle* e);
@@ -172,7 +173,7 @@ class LRUCache {
 
   // mutex_ protects the following state.
   port::Mutex mutex_;
-  size_t usage_;
+  std::atomic_size_t usage_;
 
   // Dummy head of LRU list.
   // lru.prev is newest entry, lru.next is oldest entry.
@@ -214,7 +215,7 @@ void LRUCache::FreeEntry(LRUHandle* e) {
 void LRUCache::LRU_Remove(LRUHandle* e) {
   e->next->prev = e->prev;
   e->prev->next = e->next;
-  usage_ -= e->charge;
+  usage_.fetch_sub(e->charge, std::memory_order_relaxed);
 }
 
 void LRUCache::LRU_Append(LRUHandle* e) {
@@ -223,7 +224,7 @@ void LRUCache::LRU_Append(LRUHandle* e) {
   e->prev = lru_.prev;
   e->prev->next = e;
   e->next->prev = e;
-  usage_ += e->charge;
+  usage_.fetch_add(e->charge, std::memory_order_relaxed);
 }
 
 Cache::Handle* LRUCache::Lookup(const Slice& key, uint32_t hash) {
@@ -282,7 +283,7 @@ Cache::Handle* LRUCache::Insert(
       // referenced by the cache first.
       LRUHandle* cur = lru_.next;
       for (unsigned int scanCount = 0;
-           usage_ > capacity_ && cur != &lru_
+           GetUsage() > capacity_ && cur != &lru_
            && scanCount < remove_scan_count_limit_; scanCount++) {
         LRUHandle* next = cur->next;
         if (cur->refs <= 1) {
@@ -298,7 +299,7 @@ Cache::Handle* LRUCache::Insert(
 
     // Free the space following strict LRU policy until enough space
     // is freed.
-    while (usage_ > capacity_ && lru_.next != &lru_) {
+    while (GetUsage() > capacity_ && lru_.next != &lru_) {
       LRUHandle* old = lru_.next;
       LRU_Remove(old);
       table_.Remove(old->key(), old->hash);
@@ -340,10 +341,10 @@ static int kRemoveScanCountLimit = 0; // default values, can be overridden
 
 class ShardedLRUCache : public Cache {
  private:
-  LRUCache* shard_;
+  LRUCache* shards_;
   port::Mutex id_mutex_;
   uint64_t last_id_;
-  int numShardBits;
+  int num_shard_bits_;
   size_t capacity_;
 
   static inline uint32_t HashSlice(const Slice& s) {
@@ -352,18 +353,18 @@ class ShardedLRUCache : public Cache {
 
   uint32_t Shard(uint32_t hash) {
     // Note, hash >> 32 yields hash in gcc, not the zero we expect!
-    return (numShardBits > 0) ? (hash >> (32 - numShardBits)) : 0;
+    return (num_shard_bits_ > 0) ? (hash >> (32 - num_shard_bits_)) : 0;
   }
 
   void init(size_t capacity, int numbits, int removeScanCountLimit) {
-    numShardBits = numbits;
+    num_shard_bits_ = numbits;
     capacity_ = capacity;
-    int numShards = 1 << numShardBits;
+    int num_shards = 1 << num_shard_bits_;
-    shard_ = new LRUCache[numShards];
+    shards_ = new LRUCache[num_shards];
-    const size_t per_shard = (capacity + (numShards - 1)) / numShards;
+    const size_t per_shard = (capacity + (num_shards - 1)) / num_shards;
-    for (int s = 0; s < numShards; s++) {
+    for (int s = 0; s < num_shards; s++) {
-      shard_[s].SetCapacity(per_shard);
+      shards_[s].SetCapacity(per_shard);
-      shard_[s].SetRemoveScanCountLimit(removeScanCountLimit);
+      shards_[s].SetRemoveScanCountLimit(removeScanCountLimit);
     }
   }
 
@@ -372,30 +373,30 @@ class ShardedLRUCache : public Cache {
       : last_id_(0) {
     init(capacity, kNumShardBits, kRemoveScanCountLimit);
   }
-  ShardedLRUCache(size_t capacity, int numShardBits,
+  ShardedLRUCache(size_t capacity, int num_shard_bits,
                   int removeScanCountLimit)
      : last_id_(0) {
-    init(capacity, numShardBits, removeScanCountLimit);
+    init(capacity, num_shard_bits, removeScanCountLimit);
   }
   virtual ~ShardedLRUCache() {
-    delete[] shard_;
+    delete[] shards_;
   }
   virtual Handle* Insert(const Slice& key, void* value, size_t charge,
                          void (*deleter)(const Slice& key, void* value)) {
     const uint32_t hash = HashSlice(key);
-    return shard_[Shard(hash)].Insert(key, hash, value, charge, deleter);
+    return shards_[Shard(hash)].Insert(key, hash, value, charge, deleter);
   }
   virtual Handle* Lookup(const Slice& key) {
     const uint32_t hash = HashSlice(key);
-    return shard_[Shard(hash)].Lookup(key, hash);
+    return shards_[Shard(hash)].Lookup(key, hash);
   }
   virtual void Release(Handle* handle) {
     LRUHandle* h = reinterpret_cast<LRUHandle*>(handle);
-    shard_[Shard(h->hash)].Release(handle);
+    shards_[Shard(h->hash)].Release(handle);
   }
   virtual void Erase(const Slice& key) {
     const uint32_t hash = HashSlice(key);
-    shard_[Shard(hash)].Erase(key, hash);
+    shards_[Shard(hash)].Erase(key, hash);
   }
   virtual void* Value(Handle* handle) {
     return reinterpret_cast<LRUHandle*>(handle)->value;
@@ -407,6 +408,16 @@ class ShardedLRUCache : public Cache {
   virtual size_t GetCapacity() const {
     return capacity_;
   }
+  virtual size_t GetUsage() const {
+    // We will not lock the cache when getting the usage from shards.
+    // for (size_t i = 0; i < num_shard_bits_; ++i)
+    int num_shards = 1 << num_shard_bits_;
+    size_t usage = 0;
+    for (int s = 0; s < num_shards; s++) {
+      usage += shards_[s].GetUsage();
+    }
+    return usage;
+  }
 };
 
 }  // end anonymous namespace
@@ -415,17 +426,17 @@ shared_ptr<Cache> NewLRUCache(size_t capacity) {
   return NewLRUCache(capacity, kNumShardBits);
 }
 
-shared_ptr<Cache> NewLRUCache(size_t capacity, int numShardBits) {
+shared_ptr<Cache> NewLRUCache(size_t capacity, int num_shard_bits) {
-  return NewLRUCache(capacity, numShardBits, kRemoveScanCountLimit);
+  return NewLRUCache(capacity, num_shard_bits, kRemoveScanCountLimit);
 }
 
-shared_ptr<Cache> NewLRUCache(size_t capacity, int numShardBits,
+shared_ptr<Cache> NewLRUCache(size_t capacity, int num_shard_bits,
                               int removeScanCountLimit) {
-  if (numShardBits >= 20) {
+  if (num_shard_bits >= 20) {
     return nullptr;  // the cache cannot be sharded into too many fine pieces
   }
   return std::make_shared<ShardedLRUCache>(capacity,
-                                           numShardBits,
+                                           num_shard_bits,
                                            removeScanCountLimit);
 }
 

util/cache_test.cc

@@ -107,6 +107,39 @@ class CacheTest {
 };
 CacheTest* CacheTest::current_;
 
+void dumbDeleter(const Slice& key, void* value) { }
+
+TEST(CacheTest, UsageTest) {
+  // cache is shared_ptr and will be automatically cleaned up.
+  const uint64_t kCapacity = 100000;
+  auto cache = NewLRUCache(kCapacity, 8, 200);
+
+  size_t usage = 0;
+  const char* value = "abcdef";
+  // make sure everything will be cached
+  for (int i = 1; i < 100; ++i) {
+    std::string key(i, 'a');
+    auto kv_size = key.size() + 5;
+    cache->Release(
+        cache->Insert(key, (void*)value, kv_size, dumbDeleter)
+    );
+    usage += kv_size;
+    ASSERT_EQ(usage, cache->GetUsage());
+  }
+
+  // make sure the cache will be overloaded
+  for (int i = 1; i < kCapacity; ++i) {
+    auto key = std::to_string(i);
+    cache->Release(
+        cache->Insert(key, (void*)value, key.size() + 5, dumbDeleter)
+    );
+  }
+
+  // the usage should be close to the capacity
+  ASSERT_GT(kCapacity, cache->GetUsage());
+  ASSERT_LT(kCapacity * 0.95, cache->GetUsage());
+}
+
 TEST(CacheTest, HitAndMiss) {
   ASSERT_EQ(-1, Lookup(100));
 
@@ -353,7 +386,6 @@ void deleter(const Slice& key, void* value) {
   delete (Value *)value;
 }
 
-
 TEST(CacheTest, BadEviction) {
   int n = 10;