[RocksDB][Performance Branch] Make height and branching factor configurable for skiplist implementation

Summary: As title. Especially, HashSkipListRepFactory will be able to specify a relatively small height, to reduce the memory overhead of one skiplist per bucket.

Test Plan: make check and test it on leaf4

Reviewers: dhruba, sdong, kailiu

CC: reconnect.grayhat, leveldb

Differential Revision: https://reviews.facebook.net/D14307

db/prefix_test.cc

@@ -22,6 +22,7 @@ DEFINE_uint64(items_per_prefix, 10, "total number of values per prefix");
 DEFINE_int64(write_buffer_size, 1000000000, "");
 DEFINE_int64(max_write_buffer_number, 8, "");
 DEFINE_int64(min_write_buffer_number_to_merge, 7, "");
+DEFINE_int32(skiplist_height, 4, "");
 
 // Path to the database on file system
 const std::string kDbName = rocksdb::test::TmpDir() + "/prefix_test";
@@ -111,7 +112,8 @@ class PrefixTest {
       options.prefix_extractor = prefix_extractor;
       if (FLAGS_use_nolock_version) {
         options.memtable_factory.reset(NewHashSkipListRepFactory(
-            prefix_extractor, FLAGS_bucket_count));
+                                         prefix_extractor, FLAGS_bucket_count,
+                                         FLAGS_skiplist_height));
       } else {
         options.memtable_factory =
           std::make_shared<rocksdb::PrefixHashRepFactory>(
@@ -152,7 +154,7 @@ TEST(PrefixTest, DynamicPrefixIterator) {
       TestKey test_key(prefix, sorted);
 
       Slice key = TestKeyToSlice(test_key);
-      std::string value = "v" + std::to_string(sorted);
+      std::string value(40, 0);
 
       ASSERT_OK(db->Put(write_options, key, value));
     }

db/skiplist.h

@@ -47,7 +47,8 @@ class SkipList {
   // Create a new SkipList object that will use "cmp" for comparing keys,
   // and will allocate memory using "*arena".  Objects allocated in the arena
   // must remain allocated for the lifetime of the skiplist object.
-  explicit SkipList(Comparator cmp, Arena* arena);
+  explicit SkipList(Comparator cmp, Arena* arena,
+                    int32_t max_height = 12, int32_t branching_factor = 4);
 
   // Insert key into the list.
   // REQUIRES: nothing that compares equal to key is currently in the list.
@@ -101,7 +102,8 @@ class SkipList {
   };
 
  private:
-  enum { kMaxHeight = 12 };
+  const int32_t kMaxHeight_;
+  const int32_t kBranching_;
 
   // Immutable after construction
   Comparator const compare_;
@@ -114,8 +116,8 @@ class SkipList {
   port::AtomicPointer max_height_;   // Height of the entire list
 
   // Used for optimizing sequential insert patterns
-  Node* prev_[kMaxHeight];
-  int   prev_height_;
+  Node** prev_;
+  int32_t prev_height_;
 
   inline int GetMaxHeight() const {
     return static_cast<int>(
@@ -257,13 +259,12 @@ inline void SkipList<Key,Comparator>::Iterator::SeekToLast() {
 template<typename Key, class Comparator>
 int SkipList<Key,Comparator>::RandomHeight() {
   // Increase height with probability 1 in kBranching
-  static const unsigned int kBranching = 4;
   int height = 1;
-  while (height < kMaxHeight && ((rnd_.Next() % kBranching) == 0)) {
+  while (height < kMaxHeight_ && ((rnd_.Next() % kBranching_) == 0)) {
     height++;
   }
   assert(height > 0);
-  assert(height <= kMaxHeight);
+  assert(height <= kMaxHeight_);
   return height;
 }
 
@@ -353,14 +354,24 @@ typename SkipList<Key,Comparator>::Node* SkipList<Key,Comparator>::FindLast()
 }
 
 template<typename Key, class Comparator>
-SkipList<Key,Comparator>::SkipList(Comparator cmp, Arena* arena)
-    : compare_(cmp),
+SkipList<Key,Comparator>::SkipList(Comparator cmp, Arena* arena,
+                                   int32_t max_height,
+                                   int32_t branching_factor)
+    : kMaxHeight_(max_height),
+      kBranching_(branching_factor),
+      compare_(cmp),
       arena_(arena),
-      head_(NewNode(0 /* any key will do */, kMaxHeight)),
+      head_(NewNode(0 /* any key will do */, max_height)),
       max_height_(reinterpret_cast<void*>(1)),
       prev_height_(1),
       rnd_(0xdeadbeef) {
-  for (int i = 0; i < kMaxHeight; i++) {
+  assert(kMaxHeight_ > 0);
+  assert(kBranching_ > 0);
+  // Allocate the prev_ Node* array, directly from the passed-in arena.
+  // prev_ does not need to be freed, as its life cycle is tied up with
+  // the arena as a whole.
+  prev_ = (Node**) arena_->AllocateAligned(sizeof(Node*) * kMaxHeight_);
+  for (int i = 0; i < kMaxHeight_; i++) {
     head_->SetNext(i, nullptr);
     prev_[i] = head_;
   }

include/rocksdb/memtablerep.h

@@ -267,9 +267,16 @@ public:
 
 // The same as TransformRepFactory except it doesn't use locks.
 // Experimental, will replace TransformRepFactory once we are sure
-// it performs better
+// it performs better. It contains a fixed array of buckets, each
+// pointing to a skiplist (null if the bucket is empty).
+// bucket_count: number of fixed array buckets
+// skiplist_height: the max height of the skiplist
+// skiplist_branching_factor: probabilistic size ratio between adjacent
+//                            link lists in the skiplist
 extern MemTableRepFactory* NewHashSkipListRepFactory(
-    const SliceTransform* transform, size_t bucket_count = 1000000);
+  const SliceTransform* transform, size_t bucket_count = 1000000,
+  int32_t skiplist_height = 4, int32_t skiplist_branching_factor = 4
+);
 
 }
 

util/hash_skiplist_rep.cc

@@ -20,7 +20,8 @@ namespace {
 class HashSkipListRep : public MemTableRep {
  public:
   HashSkipListRep(MemTableRep::KeyComparator& compare, Arena* arena,
-    const SliceTransform* transform, size_t bucket_size);
+                  const SliceTransform* transform, size_t bucket_size,
+                  int32_t skiplist_height, int32_t skiplist_branching_factor);
 
   virtual void Insert(const char* key) override;
 
@@ -47,6 +48,9 @@ class HashSkipListRep : public MemTableRep {
 
   size_t bucket_size_;
 
+  const int32_t skiplist_height_;
+  const int32_t skiplist_branching_factor_;
+
   // Maps slices (which are transformed user keys) to buckets of keys sharing
   // the same transform.
   port::AtomicPointer* buckets_;
@@ -215,8 +219,12 @@ class HashSkipListRep : public MemTableRep {
 };
 
 HashSkipListRep::HashSkipListRep(MemTableRep::KeyComparator& compare,
-    Arena* arena, const SliceTransform* transform, size_t bucket_size)
+                                 Arena* arena, const SliceTransform* transform,
+                                 size_t bucket_size, int32_t skiplist_height,
+                                 int32_t skiplist_branching_factor)
   : bucket_size_(bucket_size),
+    skiplist_height_(skiplist_height),
+    skiplist_branching_factor_(skiplist_branching_factor),
     transform_(transform),
     compare_(compare),
     arena_(arena),
@@ -239,7 +247,8 @@ HashSkipListRep::Bucket* HashSkipListRep::GetInitializedBucket(
   auto bucket = GetBucket(hash);
   if (bucket == nullptr) {
     auto addr = arena_->AllocateAligned(sizeof(Bucket));
-    bucket = new (addr) Bucket(compare_, arena_);
+    bucket = new (addr) Bucket(compare_, arena_, skiplist_height_,
+                               skiplist_branching_factor_);
     buckets_[hash].Release_Store(static_cast<void*>(bucket));
   }
   return bucket;
@@ -302,17 +311,23 @@ std::shared_ptr<MemTableRep::Iterator>
 
 class HashSkipListRepFactory : public MemTableRepFactory {
  public:
-  explicit HashSkipListRepFactory(const SliceTransform* transform,
-      size_t bucket_count = 1000000)
-    : transform_(transform),
-      bucket_count_(bucket_count) { }
+  explicit HashSkipListRepFactory(
+    const SliceTransform* transform,
+    size_t bucket_count,
+    int32_t skiplist_height,
+    int32_t skiplist_branching_factor)
+      : transform_(transform),
+        bucket_count_(bucket_count),
+        skiplist_height_(skiplist_height),
+        skiplist_branching_factor_(skiplist_branching_factor) { }
 
   virtual ~HashSkipListRepFactory() { delete transform_; }
 
   virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
       MemTableRep::KeyComparator& compare, Arena* arena) override {
     return std::make_shared<HashSkipListRep>(compare, arena, transform_,
-        bucket_count_);
+                                             bucket_count_, skiplist_height_,
+                                             skiplist_branching_factor_);
   }
 
   virtual const char* Name() const override {
@@ -324,11 +339,15 @@ class HashSkipListRepFactory : public MemTableRepFactory {
  private:
   const SliceTransform* transform_;
   const size_t bucket_count_;
+  const int32_t skiplist_height_;
+  const int32_t skiplist_branching_factor_;
 };
 
 MemTableRepFactory* NewHashSkipListRepFactory(
-    const SliceTransform* transform, size_t bucket_count) {
-  return new HashSkipListRepFactory(transform, bucket_count);
+    const SliceTransform* transform, size_t bucket_count,
+    int32_t skiplist_height, int32_t skiplist_branching_factor) {
+  return new HashSkipListRepFactory(transform, bucket_count,
+                                    skiplist_height, skiplist_branching_factor);
 }
 
 } // namespace rocksdb