Allow GetApproximateSize() to include mem table size if it is skip list memtable

Summary:
Add an option in GetApproximateSize() so that the result will include estimated sizes in mem tables.
To implement it, implement an estimated count from the beginning to a key in skip list. The approach is to count to find the entry, how many Next() is issued from each level, and sum them with a weight that is <branching factor> ^ <level>.

Test Plan: Add a test case

Subscribers: leveldb, dhruba

Differential Revision: https://reviews.facebook.net/D40119

HISTORY.md

@@ -15,6 +15,7 @@
 * WBWIIterator::Entry() now returns WriteEntry instead of `const WriteEntry&`
 * options.hard_rate_limit is deprecated.
 * When options.soft_rate_limit or options.level0_slowdown_writes_trigger is triggered, the way to slow down writes is changed to: write rate to DB is limited to to options.delayed_write_rate.
+* DB::GetApproximateSizes() adds a parameter to allow the estimation to include data in mem table, with default to be not to include. It is now only supported in skip list mem table.
 
 ## 3.11.0 (5/19/2015)
 ### New Features

db/db_impl.cc

@@ -3881,27 +3881,26 @@ ColumnFamilyHandle* DBImpl::GetColumnFamilyHandle(uint32_t column_family_id) {
 }
 
 void DBImpl::GetApproximateSizes(ColumnFamilyHandle* column_family,
-                                 const Range* range, int n, uint64_t* sizes) {
+                                 const Range* range, int n, uint64_t* sizes,
+                                 bool include_memtable) {
   Version* v;
   auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
   auto cfd = cfh->cfd();
-  {
-    InstrumentedMutexLock l(&mutex_);
-    v = cfd->current();
-    v->Ref();
-  }
+  SuperVersion* sv = GetAndRefSuperVersion(cfd);
+  v = sv->current;
 
   for (int i = 0; i < n; i++) {
     // Convert user_key into a corresponding internal key.
     InternalKey k1(range[i].start, kMaxSequenceNumber, kValueTypeForSeek);
     InternalKey k2(range[i].limit, kMaxSequenceNumber, kValueTypeForSeek);
     sizes[i] = versions_->ApproximateSize(v, k1.Encode(), k2.Encode());
+    if (include_memtable) {
+      sizes[i] += sv->mem->ApproximateSize(k1.Encode(), k2.Encode());
+      sizes[i] += sv->imm->ApproximateSize(k1.Encode(), k2.Encode());
+    }
   }
 
-  {
-    InstrumentedMutexLock l(&mutex_);
-    v->Unref();
-  }
+  ReturnAndCleanupSuperVersion(cfd, sv);
 }
 
 std::list<uint64_t>::iterator

db/db_impl.h

@@ -122,8 +122,8 @@ class DBImpl : public DB {
                               const Slice& property, uint64_t* value) override;
   using DB::GetApproximateSizes;
   virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
-                                   const Range* range, int n,
-                                   uint64_t* sizes) override;
+                                   const Range* range, int n, uint64_t* sizes,
+                                   bool include_memtable = false) override;
   using DB::CompactRange;
   virtual Status CompactRange(ColumnFamilyHandle* column_family,
                               const Slice* begin, const Slice* end,

db/db_test.cc

@@ -6577,6 +6577,112 @@ static bool Between(uint64_t val, uint64_t low, uint64_t high) {
   return result;
 }
 
+TEST_F(DBTest, ApproximateSizesMemTable) {
+  Options options;
+  options.write_buffer_size = 100000000;  // Large write buffer
+  options.compression = kNoCompression;
+  options.create_if_missing = true;
+  options = CurrentOptions(options);
+  DestroyAndReopen(options);
+
+  const int N = 128;
+  Random rnd(301);
+  for (int i = 0; i < N; i++) {
+    ASSERT_OK(Put(Key(i), RandomString(&rnd, 1024)));
+  }
+
+  uint64_t size;
+  std::string start = Key(50);
+  std::string end = Key(60);
+  Range r(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_GT(size, 6000);
+  ASSERT_LT(size, 204800);
+  // Zero if not including mem table
+  db_->GetApproximateSizes(&r, 1, &size, false);
+  ASSERT_EQ(size, 0);
+
+  start = Key(500);
+  end = Key(600);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_EQ(size, 0);
+
+  for (int i = 0; i < N; i++) {
+    ASSERT_OK(Put(Key(1000 + i), RandomString(&rnd, 1024)));
+  }
+
+  start = Key(500);
+  end = Key(600);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_EQ(size, 0);
+
+  start = Key(100);
+  end = Key(1020);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_GT(size, 6000);
+
+  options.max_write_buffer_number = 8;
+  options.min_write_buffer_number_to_merge = 5;
+  options.write_buffer_size = 1024 * N;  // Not very large
+  DestroyAndReopen(options);
+
+  int keys[N * 3];
+  for (int i = 0; i < N; i++) {
+    keys[i * 3] = i * 5;
+    keys[i * 3 + 1] = i * 5 + 1;
+    keys[i * 3 + 2] = i * 5 + 2;
+  }
+  std::random_shuffle(std::begin(keys), std::end(keys));
+
+  for (int i = 0; i < N * 3; i++) {
+    ASSERT_OK(Put(Key(keys[i] + 1000), RandomString(&rnd, 1024)));
+  }
+
+  start = Key(100);
+  end = Key(300);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_EQ(size, 0);
+
+  start = Key(1050);
+  end = Key(1080);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_GT(size, 6000);
+
+  start = Key(2100);
+  end = Key(2300);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size, true);
+  ASSERT_EQ(size, 0);
+
+  start = Key(1050);
+  end = Key(1080);
+  r = Range(start, end);
+  uint64_t size_with_mt, size_without_mt;
+  db_->GetApproximateSizes(&r, 1, &size_with_mt, true);
+  ASSERT_GT(size_with_mt, 6000);
+  db_->GetApproximateSizes(&r, 1, &size_without_mt, false);
+  ASSERT_EQ(size_without_mt, 0);
+
+  Flush();
+
+  for (int i = 0; i < N; i++) {
+    ASSERT_OK(Put(Key(i + 1000), RandomString(&rnd, 1024)));
+  }
+
+  start = Key(1050);
+  end = Key(1080);
+  r = Range(start, end);
+  db_->GetApproximateSizes(&r, 1, &size_with_mt, true);
+  db_->GetApproximateSizes(&r, 1, &size_without_mt, false);
+  ASSERT_GT(size_with_mt, size_without_mt);
+  ASSERT_GT(size_without_mt, 6000);
+}
+
 TEST_F(DBTest, ApproximateSizes) {
   do {
     Options options;
@@ -8948,8 +9054,8 @@ class ModelDB: public DB {
   }
   using DB::GetApproximateSizes;
   virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
-                                   const Range* range, int n,
-                                   uint64_t* sizes) override {
+                                   const Range* range, int n, uint64_t* sizes,
+                                   bool include_memtable) override {
     for (int i = 0; i < n; i++) {
       sizes[i] = 0;
     }

db/memtable.cc

@@ -64,6 +64,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
       table_(ioptions.memtable_factory->CreateMemTableRep(
           comparator_, &allocator_, ioptions.prefix_extractor,
           ioptions.info_log)),
+      data_size_(0),
       num_entries_(0),
       num_deletes_(0),
       flush_in_progress_(false),
@@ -290,6 +291,26 @@ port::RWMutex* MemTable::GetLock(const Slice& key) {
   return &locks_[hash(key) % locks_.size()];
 }
 
+uint64_t MemTable::ApproximateSize(const Slice& start_ikey,
+                                   const Slice& end_ikey) {
+  uint64_t entry_count = table_->ApproximateNumEntries(start_ikey, end_ikey);
+  if (entry_count == 0) {
+    return 0;
+  }
+  uint64_t n = num_entries_.load(std::memory_order_relaxed);
+  if (n == 0) {
+    return 0;
+  }
+  if (entry_count > n) {
+    // table_->ApproximateNumEntries() is just an estimate so it can be larger
+    // than actual entries we have. Cap it to entries we have to limit the
+    // inaccuracy.
+    entry_count = n;
+  }
+  uint64_t data_size = data_size_.load(std::memory_order_relaxed);
+  return entry_count * (data_size / n);
+}
+
 void MemTable::Add(SequenceNumber s, ValueType type,
                    const Slice& key, /* user key */
                    const Slice& value) {
@@ -317,7 +338,10 @@ void MemTable::Add(SequenceNumber s, ValueType type,
   memcpy(p, value.data(), val_size);
   assert((unsigned)(p + val_size - buf) == (unsigned)encoded_len);
   table_->Insert(handle);
-  num_entries_++;
+  num_entries_.store(num_entries_.load(std::memory_order_relaxed) + 1,
+                     std::memory_order_relaxed);
+  data_size_.store(data_size_.load(std::memory_order_relaxed) + encoded_len,
+                   std::memory_order_relaxed);
   if (type == kTypeDeletion) {
     num_deletes_++;
   }

db/memtable.h

@@ -212,7 +212,9 @@ class MemTable {
   // Get total number of entries in the mem table.
   // REQUIRES: external synchronization to prevent simultaneous
   // operations on the same MemTable (unless this Memtable is immutable).
-  uint64_t num_entries() const { return num_entries_; }
+  uint64_t num_entries() const {
+    return num_entries_.load(std::memory_order_relaxed);
+  }
 
   // Get total number of deletes in the mem table.
   // REQUIRES: external synchronization to prevent simultaneous
@@ -275,6 +277,8 @@ class MemTable {
     return table_->IsSnapshotSupported() && !moptions_.inplace_update_support;
   }
 
+  uint64_t ApproximateSize(const Slice& start_ikey, const Slice& end_ikey);
+
   // Get the lock associated for the key
   port::RWMutex* GetLock(const Slice& key);
 
@@ -300,7 +304,9 @@ class MemTable {
   MemTableAllocator allocator_;
   unique_ptr<MemTableRep> table_;
 
-  uint64_t num_entries_;
+  // Total data size of all data inserted
+  std::atomic<uint64_t> data_size_;
+  std::atomic<uint64_t> num_entries_;
   uint64_t num_deletes_;
 
   // These are used to manage memtable flushes to storage

db/memtable_list.cc

@@ -149,6 +149,15 @@ uint64_t MemTableListVersion::GetTotalNumEntries() const {
   return total_num;
 }
 
+uint64_t MemTableListVersion::ApproximateSize(const Slice& start_ikey,
+                                              const Slice& end_ikey) {
+  uint64_t total_size = 0;
+  for (auto& m : memlist_) {
+    total_size += m->ApproximateSize(start_ikey, end_ikey);
+  }
+  return total_size;
+}
+
 uint64_t MemTableListVersion::GetTotalNumDeletes() const {
   uint64_t total_num = 0;
   for (auto& m : memlist_) {

db/memtable_list.h

@@ -84,6 +84,8 @@ class MemTableListVersion {
 
   uint64_t GetTotalNumDeletes() const;
 
+  uint64_t ApproximateSize(const Slice& start_ikey, const Slice& end_ikey);
+
   // Returns the value of MemTable::GetEarliestSequenceNumber() on the most
   // recent MemTable in this list or kMaxSequenceNumber if the list is empty.
   // If include_history=true, will also search Memtables in MemTableList

db/skiplist.h

@@ -59,6 +59,9 @@ class SkipList {
   // Returns true iff an entry that compares equal to key is in the list.
   bool Contains(const Key& key) const;
 
+  // Return estimated number of entries smaller than `key`.
+  uint64_t EstimateCount(const Key& key) const;
+
   // Iteration over the contents of a skip list
   class Iterator {
    public:
@@ -354,10 +357,34 @@ typename SkipList<Key, Comparator>::Node* SkipList<Key, Comparator>::FindLast()
   }
 }
 
-template<typename Key, class Comparator>
+template <typename Key, class Comparator>
+uint64_t SkipList<Key, Comparator>::EstimateCount(const Key& key) const {
+  uint64_t count = 0;
+
+  Node* x = head_;
+  int level = GetMaxHeight() - 1;
+  while (true) {
+    assert(x == head_ || compare_(x->key, key) < 0);
+    Node* next = x->Next(level);
+    if (next == nullptr || compare_(next->key, key) >= 0) {
+      if (level == 0) {
+        return count;
+      } else {
+        // Switch to next list
+        count *= kBranching_;
+        level--;
+      }
+    } else {
+      x = next;
+      count++;
+    }
+  }
+}
+
+template <typename Key, class Comparator>
 SkipList<Key, Comparator>::SkipList(const Comparator cmp, Allocator* allocator,
-                                   int32_t max_height,
-                                   int32_t branching_factor)
+                                    int32_t max_height,
+                                    int32_t branching_factor)
     : kMaxHeight_(max_height),
       kBranching_(branching_factor),
       compare_(cmp),

include/rocksdb/db.h

@@ -397,10 +397,12 @@ class DB {
   //
   // The results may not include the sizes of recently written data.
   virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
-                                   const Range* range, int n,
-                                   uint64_t* sizes) = 0;
-  virtual void GetApproximateSizes(const Range* range, int n, uint64_t* sizes) {
-    GetApproximateSizes(DefaultColumnFamily(), range, n, sizes);
+                                   const Range* range, int n, uint64_t* sizes,
+                                   bool include_memtable = false) = 0;
+  virtual void GetApproximateSizes(const Range* range, int n, uint64_t* sizes,
+                                   bool include_memtable = false) {
+    GetApproximateSizes(DefaultColumnFamily(), range, n, sizes,
+                        include_memtable);
   }
 
   // Compact the underlying storage for the key range [*begin,*end].

include/rocksdb/memtablerep.h

@@ -103,6 +103,11 @@ class MemTableRep {
   virtual void Get(const LookupKey& k, void* callback_args,
                    bool (*callback_func)(void* arg, const char* entry));
 
+  virtual uint64_t ApproximateNumEntries(const Slice& start_ikey,
+                                         const Slice& end_key) {
+    return 0;
+  }
+
   // Report an approximation of how much memory has been used other than memory
   // that was allocated through the allocator.
   virtual size_t ApproximateMemoryUsage() = 0;

include/rocksdb/utilities/stackable_db.h

@@ -121,8 +121,8 @@ class StackableDB : public DB {
 
   using DB::GetApproximateSizes;
   virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
-                                   const Range* r, int n,
-                                   uint64_t* sizes) override {
+                                   const Range* r, int n, uint64_t* sizes,
+                                   bool include_memtable = false) override {
       return db_->GetApproximateSizes(column_family, r, n, sizes);
   }
 

util/skiplistrep.cc

@@ -52,6 +52,15 @@ public:
     }
   }
 
+  uint64_t ApproximateNumEntries(const Slice& start_ikey,
+                                 const Slice& end_ikey) override {
+    std::string tmp;
+    uint64_t start_count =
+        skip_list_.EstimateCount(EncodeKey(&tmp, start_ikey));
+    uint64_t end_count = skip_list_.EstimateCount(EncodeKey(&tmp, end_ikey));
+    return (end_count >= start_count) ? (end_count - start_count) : 0;
+  }
+
   virtual ~SkipListRep() override { }
 
   // Iteration over the contents of a skip list