Summary: This diff adds support for concurrent adds to the skiplist memtable implementations. Memory allocation is made thread-safe by the addition of a spinlock, with small per-core buffers to avoid contention. Concurrent memtable writes are made via an additional method and don't impose a performance overhead on the non-concurrent case, so parallelism can be selected on a per-batch basis. Write thread synchronization is an increasing bottleneck for higher levels of concurrency, so this diff adds --enable_write_thread_adaptive_yield (default off). This feature causes threads joining a write batch group to spin for a short time (default 100 usec) using sched_yield, rather than going to sleep on a mutex. If the timing of the yield calls indicates that another thread has actually run during the yield then spinning is avoided. This option improves performance for concurrent situations even without parallel adds, although it has the potential to increase CPU usage (and the heuristic adaptation is not yet mature). Parallel writes are not currently compatible with inplace updates, update callbacks, or delete filtering. Enable it with --allow_concurrent_memtable_write (and --enable_write_thread_adaptive_yield). Parallel memtable writes are performance neutral when there is no actual parallelism, and in my experiments (SSD server-class Linux and varying contention and key sizes for fillrandom) they are always a performance win when there is more than one thread. Statistics are updated earlier in the write path, dropping the number of DB mutex acquisitions from 2 to 1 for almost all cases. This diff was motivated and inspired by Yahoo's cLSM work. It is more conservative than cLSM: RocksDB's write batch group leader role is preserved (along with all of the existing flush and write throttling logic) and concurrent writers are blocked until all memtable insertions have completed and the sequence number has been advanced, to preserve linearizability. My test config is "db_bench -benchmarks=fillrandom -threads=$T -batch_size=1 -memtablerep=skip_list -value_size=100 --num=1000000/$T -level0_slowdown_writes_trigger=9999 -level0_stop_writes_trigger=9999 -disable_auto_compactions --max_write_buffer_number=8 -max_background_flushes=8 --disable_wal --write_buffer_size=160000000 --block_size=16384 --allow_concurrent_memtable_write" on a two-socket Xeon E5-2660 @ 2.2Ghz with lots of memory and an SSD hard drive. With 1 thread I get ~440Kops/sec. Peak performance for 1 socket (numactl -N1) is slightly more than 1Mops/sec, at 16 threads. Peak performance across both sockets happens at 30 threads, and is ~900Kops/sec, although with fewer threads there is less performance loss when the system has background work. Test Plan: 1. concurrent stress tests for InlineSkipList and DynamicBloom 2. make clean; make check 3. make clean; DISABLE_JEMALLOC=1 make valgrind_check; valgrind db_bench 4. make clean; COMPILE_WITH_TSAN=1 make all check; db_bench 5. make clean; COMPILE_WITH_ASAN=1 make all check; db_bench 6. make clean; OPT=-DROCKSDB_LITE make check 7. verify no perf regressions when disabled Reviewers: igor, sdong Reviewed By: sdong Subscribers: MarkCallaghan, IslamAbdelRahman, anthony, yhchiang, rven, sdong, guyg8, kradhakrishnan, dhruba Differential Revision: https://reviews.facebook.net/D50589main
Nathan Bronson
9 years ago
parent
5b2587b5cb
commit
7d87f02799
@ -0,0 +1,49 @@ |
||||
// Copyright (c) 2013, Facebook, Inc. All rights reserved.
|
||||
// This source code is licensed under the BSD-style license found in the
|
||||
// LICENSE file in the root directory of this source tree. An additional grant
|
||||
// of patent rights can be found in the PATENTS file in the same 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.
|
||||
|
||||
#include "util/concurrent_arena.h" |
||||
#include <thread> |
||||
#include "port/likely.h" |
||||
#include "port/port.h" |
||||
#include "util/random.h" |
||||
|
||||
namespace rocksdb { |
||||
|
||||
#if ROCKSDB_SUPPORT_THREAD_LOCAL |
||||
__thread uint32_t ConcurrentArena::tls_cpuid = 0; |
||||
#endif |
||||
|
||||
ConcurrentArena::ConcurrentArena(size_t block_size, size_t huge_page_size) |
||||
: shard_block_size_(block_size / 8), arena_(block_size, huge_page_size) { |
||||
// find a power of two >= num_cpus and >= 8
|
||||
auto num_cpus = std::thread::hardware_concurrency(); |
||||
index_mask_ = 7; |
||||
while (index_mask_ + 1 < num_cpus) { |
||||
index_mask_ = index_mask_ * 2 + 1; |
||||
} |
||||
|
||||
shards_.reset(new Shard[index_mask_ + 1]); |
||||
Fixup(); |
||||
} |
||||
|
||||
ConcurrentArena::Shard* ConcurrentArena::Repick() { |
||||
int cpuid = port::PhysicalCoreID(); |
||||
if (UNLIKELY(cpuid < 0)) { |
||||
// cpu id unavailable, just pick randomly
|
||||
cpuid = Random::GetTLSInstance()->Uniform(index_mask_ + 1); |
||||
} |
||||
#if ROCKSDB_SUPPORT_THREAD_LOCAL |
||||
// even if we are cpu 0, use a non-zero tls_cpuid so we can tell we
|
||||
// have repicked
|
||||
tls_cpuid = cpuid | (index_mask_ + 1); |
||||
#endif |
||||
return &shards_[cpuid & index_mask_]; |
||||
} |
||||
|
||||
} // namespace rocksdb
|
||||
@ -0,0 +1,192 @@ |
||||
// Copyright (c) 2013, Facebook, Inc. All rights reserved.
|
||||
// This source code is licensed under the BSD-style license found in the
|
||||
// LICENSE file in the root directory of this source tree. An additional grant
|
||||
// of patent rights can be found in the PATENTS file in the same 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.
|
||||
|
||||
#pragma once |
||||
#include <atomic> |
||||
#include <memory> |
||||
#include <utility> |
||||
#include "port/likely.h" |
||||
#include "util/allocator.h" |
||||
#include "util/arena.h" |
||||
#include "util/mutexlock.h" |
||||
#include "util/thread_local.h" |
||||
|
||||
namespace rocksdb { |
||||
|
||||
class Logger; |
||||
|
||||
// ConcurrentArena wraps an Arena. It makes it thread safe using a fast
|
||||
// inlined spinlock, and adds small per-core allocation caches to avoid
|
||||
// contention for small allocations. To avoid any memory waste from the
|
||||
// per-core shards, they are kept small, they are lazily instantiated
|
||||
// only if ConcurrentArena actually notices concurrent use, and they
|
||||
// adjust their size so that there is no fragmentation waste when the
|
||||
// shard blocks are allocated from the underlying main arena.
|
||||
class ConcurrentArena : public Allocator { |
||||
public: |
||||
// block_size and huge_page_size are the same as for Arena (and are
|
||||
// in fact just passed to the constructor of arena_. The core-local
|
||||
// shards compute their shard_block_size as a fraction of block_size
|
||||
// that varies according to the hardware concurrency level.
|
||||
explicit ConcurrentArena(size_t block_size = Arena::kMinBlockSize, |
||||
size_t huge_page_size = 0); |
||||
|
||||
char* Allocate(size_t bytes) override { |
||||
return AllocateImpl(bytes, false /*force_arena*/, |
||||
[=]() { return arena_.Allocate(bytes); }); |
||||
} |
||||
|
||||
char* AllocateAligned(size_t bytes, size_t huge_page_size = 0, |
||||
Logger* logger = nullptr) override { |
||||
size_t rounded_up = ((bytes - 1) | (sizeof(void*) - 1)) + 1; |
||||
assert(rounded_up >= bytes && rounded_up < bytes + sizeof(void*) && |
||||
(rounded_up % sizeof(void*)) == 0); |
||||
|
||||
return AllocateImpl(rounded_up, huge_page_size != 0 /*force_arena*/, [=]() { |
||||
return arena_.AllocateAligned(rounded_up, huge_page_size, logger); |
||||
}); |
||||
} |
||||
|
||||
size_t ApproximateMemoryUsage() const { |
||||
std::unique_lock<SpinMutex> lock(arena_mutex_, std::defer_lock); |
||||
if (index_mask_ != 0) { |
||||
lock.lock(); |
||||
} |
||||
return arena_.ApproximateMemoryUsage() - ShardAllocatedAndUnused(); |
||||
} |
||||
|
||||
size_t MemoryAllocatedBytes() const { |
||||
return memory_allocated_bytes_.load(std::memory_order_relaxed); |
||||
} |
||||
|
||||
size_t AllocatedAndUnused() const { |
||||
return arena_allocated_and_unused_.load(std::memory_order_relaxed) + |
||||
ShardAllocatedAndUnused(); |
||||
} |
||||
|
||||
size_t IrregularBlockNum() const { |
||||
return irregular_block_num_.load(std::memory_order_relaxed); |
||||
} |
||||
|
||||
size_t BlockSize() const override { return arena_.BlockSize(); } |
||||
|
||||
private: |
||||
struct Shard { |
||||
char padding[40]; |
||||
mutable SpinMutex mutex; |
||||
char* free_begin_; |
||||
std::atomic<size_t> allocated_and_unused_; |
||||
|
||||
Shard() : allocated_and_unused_(0) {} |
||||
}; |
||||
|
||||
#if ROCKSDB_SUPPORT_THREAD_LOCAL |
||||
static __thread uint32_t tls_cpuid; |
||||
#else |
||||
enum ZeroFirstEnum : uint32_t { tls_cpuid = 0 }; |
||||
#endif |
||||
|
||||
char padding0[56]; |
||||
|
||||
size_t shard_block_size_; |
||||
|
||||
// shards_[i & index_mask_] is valid
|
||||
size_t index_mask_; |
||||
std::unique_ptr<Shard[]> shards_; |
||||
|
||||
Arena arena_; |
||||
mutable SpinMutex arena_mutex_; |
||||
std::atomic<size_t> arena_allocated_and_unused_; |
||||
std::atomic<size_t> memory_allocated_bytes_; |
||||
std::atomic<size_t> irregular_block_num_; |
||||
|
||||
char padding1[56]; |
||||
|
||||
Shard* Repick(); |
||||
|
||||
size_t ShardAllocatedAndUnused() const { |
||||
size_t total = 0; |
||||
for (size_t i = 0; i <= index_mask_; ++i) { |
||||
total += shards_[i].allocated_and_unused_.load(std::memory_order_relaxed); |
||||
} |
||||
return total; |
||||
} |
||||
|
||||
template <typename Func> |
||||
char* AllocateImpl(size_t bytes, bool force_arena, const Func& func) { |
||||
uint32_t cpu; |
||||
|
||||
// Go directly to the arena if the allocation is too large, or if
|
||||
// we've never needed to Repick() and the arena mutex is available
|
||||
// with no waiting. This keeps the fragmentation penalty of
|
||||
// concurrency zero unless it might actually confer an advantage.
|
||||
std::unique_lock<SpinMutex> arena_lock(arena_mutex_, std::defer_lock); |
||||
if (bytes > shard_block_size_ / 4 || force_arena || |
||||
((cpu = tls_cpuid) == 0 && |
||||
!shards_[0].allocated_and_unused_.load(std::memory_order_relaxed) && |
||||
arena_lock.try_lock())) { |
||||
if (!arena_lock.owns_lock()) { |
||||
arena_lock.lock(); |
||||
} |
||||
auto rv = func(); |
||||
Fixup(); |
||||
return rv; |
||||
} |
||||
|
||||
// pick a shard from which to allocate
|
||||
Shard* s = &shards_[cpu & index_mask_]; |
||||
if (!s->mutex.try_lock()) { |
||||
s = Repick(); |
||||
s->mutex.lock(); |
||||
} |
||||
std::unique_lock<SpinMutex> lock(s->mutex, std::adopt_lock); |
||||
|
||||
size_t avail = s->allocated_and_unused_.load(std::memory_order_relaxed); |
||||
if (avail < bytes) { |
||||
// reload
|
||||
std::lock_guard<SpinMutex> reload_lock(arena_mutex_); |
||||
|
||||
// If the arena's current block is within a factor of 2 of the right
|
||||
// size, we adjust our request to avoid arena waste.
|
||||
auto exact = arena_allocated_and_unused_.load(std::memory_order_relaxed); |
||||
assert(exact == arena_.AllocatedAndUnused()); |
||||
avail = exact >= shard_block_size_ / 2 && exact < shard_block_size_ * 2 |
||||
? exact |
||||
: shard_block_size_; |
||||
s->free_begin_ = arena_.AllocateAligned(avail); |
||||
Fixup(); |
||||
} |
||||
s->allocated_and_unused_.store(avail - bytes, std::memory_order_relaxed); |
||||
|
||||
char* rv; |
||||
if ((bytes % sizeof(void*)) == 0) { |
||||
// aligned allocation from the beginning
|
||||
rv = s->free_begin_; |
||||
s->free_begin_ += bytes; |
||||
} else { |
||||
// unaligned from the end
|
||||
rv = s->free_begin_ + avail - bytes; |
||||
} |
||||
return rv; |
||||
} |
||||
|
||||
void Fixup() { |
||||
arena_allocated_and_unused_.store(arena_.AllocatedAndUnused(), |
||||
std::memory_order_relaxed); |
||||
memory_allocated_bytes_.store(arena_.MemoryAllocatedBytes(), |
||||
std::memory_order_relaxed); |
||||
irregular_block_num_.store(arena_.IrregularBlockNum(), |
||||
std::memory_order_relaxed); |
||||
} |
||||
|
||||
ConcurrentArena(const ConcurrentArena&) = delete; |
||||
ConcurrentArena& operator=(const ConcurrentArena&) = delete; |
||||
}; |
||||
|
||||
} // namespace rocksdb
|
||||
Loading…
Reference in new issue