rocksdb/table/block_based/block_prefetcher.h
Akanksha Mahajan a0e0feca62 Improve BlockPrefetcher to prefetch only for sequential scans (#7394)
Summary:
BlockPrefetcher is used by iterators to prefetch data if they
anticipate more data to be used in future and this is valid for forward sequential
scans. But BlockPrefetcher tracks only num_file_reads_ and not if reads
are sequential. This presents problem for MultiGet with large number of
keys when it reseeks index iterator and data block. FilePrefetchBuffer
can end up doing large readahead for reseeks as readahead size
increases exponentially once readahead is enabled. Same issue is with
BlockBasedTableIterator.

Add previous length and offset read as well in BlockPrefetcher (creates
FilePrefetchBuffer) and FilePrefetchBuffer (does prefetching of data) to
determine if reads are sequential and then  prefetch.

Update the last block read after cache hit to take reads from cache also
in account.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/7394

Test Plan: Add new unit test case

Reviewed By: anand1976

Differential Revision: D23737617

Pulled By: akankshamahajan15

fbshipit-source-id: 8e6917c25ed87b285ee495d1b68dc623d71205a3
2021-04-28 12:53:46 -07:00

51 lines
1.8 KiB
C++

// 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.
#pragma once
#include "table/block_based/block_based_table_reader.h"
namespace ROCKSDB_NAMESPACE {
class BlockPrefetcher {
public:
explicit BlockPrefetcher(size_t compaction_readahead_size)
: compaction_readahead_size_(compaction_readahead_size) {}
void PrefetchIfNeeded(const BlockBasedTable::Rep* rep,
const BlockHandle& handle, size_t readahead_size,
bool is_for_compaction);
FilePrefetchBuffer* prefetch_buffer() { return prefetch_buffer_.get(); }
void UpdateReadPattern(const size_t& offset, const size_t& len) {
prev_offset_ = offset;
prev_len_ = len;
}
bool IsBlockSequential(const size_t& offset) {
return (prev_len_ == 0 || (prev_offset_ + prev_len_ == offset));
}
void ResetValues() {
num_file_reads_ = 1;
readahead_size_ = BlockBasedTable::kInitAutoReadaheadSize;
readahead_limit_ = 0;
return;
}
private:
// Readahead size used in compaction, its value is used only if
// lookup_context_.caller = kCompaction.
size_t compaction_readahead_size_;
size_t readahead_size_ = BlockBasedTable::kInitAutoReadaheadSize;
size_t readahead_limit_ = 0;
int64_t num_file_reads_ = 0;
size_t prev_offset_ = 0;
size_t prev_len_ = 0;
std::unique_ptr<FilePrefetchBuffer> prefetch_buffer_;
};
} // namespace ROCKSDB_NAMESPACE