Revision 2f41ecfe75f0ebf33e4969083b031c7a97ebaee7 authored by Zhongyi Xie on 23 August 2019, 20:54:09 UTC, committed by Facebook Github Bot on 23 August 2019, 20:55:34 UTC
Summary:
MyRocks currently sets `max_write_buffer_number_to_maintain` in order to maintain enough history for transaction conflict checking. The effectiveness of this approach depends on the size of memtables. When memtables are small, it may not keep enough history; when memtables are large, this may consume too much memory.
We are proposing a new way to configure memtable list history: by limiting the memory usage of immutable memtables. The new option is `max_write_buffer_size_to_maintain` and it will take precedence over the old `max_write_buffer_number_to_maintain` if they are both set to non-zero values. The new option accounts for the total memory usage of flushed immutable memtables and mutable memtable. When the total usage exceeds the limit, RocksDB may start dropping immutable memtables (which is also called trimming history), starting from the oldest one.
The semantics of the old option actually works both as an upper bound and lower bound. History trimming will start if number of immutable memtables exceeds the limit, but it will never go below (limit-1) due to history trimming.
In order the mimic the behavior with the new option, history trimming will stop if dropping the next immutable memtable causes the total memory usage go below the size limit. For example, assuming the size limit is set to 64MB, and there are 3 immutable memtables with sizes of 20, 30, 30. Although the total memory usage is 80MB > 64MB, dropping the oldest memtable will reduce the memory usage to 60MB < 64MB, so in this case no memtable will be dropped.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5022

Differential Revision: D14394062

Pulled By: miasantreble

fbshipit-source-id: 60457a509c6af89d0993f988c9b5c2aa9e45f5c5
1 parent 26293c8
Raw File
port_example.h
//  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.
//
// This file contains the specification, but not the implementations,
// of the types/operations/etc. that should be defined by a platform
// specific port_<platform>.h file.  Use this file as a reference for
// how to port this package to a new platform.

#pragma once

namespace rocksdb {
namespace port {

// TODO(jorlow): Many of these belong more in the environment class rather than
//               here. We should try moving them and see if it affects perf.

// The following boolean constant must be true on a little-endian machine
// and false otherwise.
static const bool kLittleEndian = true /* or some other expression */;

// ------------------ Threading -------------------

// A Mutex represents an exclusive lock.
class Mutex {
 public:
  Mutex();
  ~Mutex();

  // Lock the mutex.  Waits until other lockers have exited.
  // Will deadlock if the mutex is already locked by this thread.
  void Lock();

  // Unlock the mutex.
  // REQUIRES: This mutex was locked by this thread.
  void Unlock();

  // Optionally crash if this thread does not hold this mutex.
  // The implementation must be fast, especially if NDEBUG is
  // defined.  The implementation is allowed to skip all checks.
  void AssertHeld();
};

class CondVar {
 public:
  explicit CondVar(Mutex* mu);
  ~CondVar();

  // Atomically release *mu and block on this condition variable until
  // either a call to SignalAll(), or a call to Signal() that picks
  // this thread to wakeup.
  // REQUIRES: this thread holds *mu
  void Wait();

  // If there are some threads waiting, wake up at least one of them.
  void Signal();

  // Wake up all waiting threads.
  void SignallAll();
};

// Thread-safe initialization.
// Used as follows:
//      static port::OnceType init_control = LEVELDB_ONCE_INIT;
//      static void Initializer() { ... do something ...; }
//      ...
//      port::InitOnce(&init_control, &Initializer);
typedef intptr_t OnceType;
#define LEVELDB_ONCE_INIT 0
extern void InitOnce(port::OnceType*, void (*initializer)());

// ------------------ Compression -------------------

// Store the snappy compression of "input[0,input_length-1]" in *output.
// Returns false if snappy is not supported by this port.
extern bool Snappy_Compress(const char* input, size_t input_length,
                            std::string* output);

// If input[0,input_length-1] looks like a valid snappy compressed
// buffer, store the size of the uncompressed data in *result and
// return true.  Else return false.
extern bool Snappy_GetUncompressedLength(const char* input, size_t length,
                                         size_t* result);

// Attempt to snappy uncompress input[0,input_length-1] into *output.
// Returns true if successful, false if the input is invalid lightweight
// compressed data.
//
// REQUIRES: at least the first "n" bytes of output[] must be writable
// where "n" is the result of a successful call to
// Snappy_GetUncompressedLength.
extern bool Snappy_Uncompress(const char* input_data, size_t input_length,
                              char* output);

}  // namespace port
}  // namespace rocksdb
back to top