rocksdb/table/block_based/index_reader_common.h

92 lines
3.3 KiB
C
Raw Normal View History

// 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"
#include "table/block_based/reader_common.h"
namespace ROCKSDB_NAMESPACE {
// Encapsulates common functionality for the various index reader
// implementations. Provides access to the index block regardless of whether
// it is owned by the reader or stored in the cache, or whether it is pinned
// in the cache or not.
class BlockBasedTable::IndexReaderCommon : public BlockBasedTable::IndexReader {
public:
IndexReaderCommon(const BlockBasedTable* t,
CachableEntry<Block>&& index_block)
: table_(t), index_block_(std::move(index_block)) {
assert(table_ != nullptr);
}
protected:
static Status ReadIndexBlock(const BlockBasedTable* table,
FilePrefetchBuffer* prefetch_buffer,
const ReadOptions& read_options, bool use_cache,
GetContext* get_context,
BlockCacheLookupContext* lookup_context,
CachableEntry<Block>* index_block);
const BlockBasedTable* table() const { return table_; }
const InternalKeyComparator* internal_comparator() const {
assert(table_ != nullptr);
assert(table_->get_rep() != nullptr);
return &table_->get_rep()->internal_comparator;
}
bool index_has_first_key() const {
assert(table_ != nullptr);
assert(table_->get_rep() != nullptr);
return table_->get_rep()->index_has_first_key;
}
bool index_key_includes_seq() const {
assert(table_ != nullptr);
assert(table_->get_rep() != nullptr);
return table_->get_rep()->index_key_includes_seq;
}
bool index_value_is_full() const {
assert(table_ != nullptr);
assert(table_->get_rep() != nullptr);
return table_->get_rep()->index_value_is_full;
}
bool cache_index_blocks() const {
assert(table_ != nullptr);
assert(table_->get_rep() != nullptr);
return table_->get_rep()->table_options.cache_index_and_filter_blocks;
}
Add support to strip / pad timestamp when creating / reading a block based table (#11495) Summary: Add support to strip timestamp in block based table builder and pad timestamp in block based table reader. On the write path, use the per column family option `AdvancedColumnFamilyOptions.persist_user_defined_timestamps` to indicate whether user-defined timestamps should be stripped for all block based tables created for the column family. On the read path, added a per table `TableReadOption.user_defined_timestamps_persisted` to flag whether the user keys in the table contains user defined timestamps. This patch is mostly passing the related flags down to the block building/parsing level with the exception of handling the `first_internal_key` in `IndexValue`, which is included in the `IndexBuilder` level. The value part of range deletion entries should have a similar handling, I haven't decided where to best fit this piece of logic, I will do it in a follow up. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11495 Test Plan: Existing test `BlockBasedTableReaderTest` is parameterized to run with: 1) different UDT test modes: kNone, kNormal, kStripUserDefinedTimestamp 2) all four index types, when index type is `kTwoLevelIndexSearch`, also enables partitioned filters 3) parallel vs non-parallel compression 4) enable/disable compression dictionary. Also added tests for API `BlockBasedTableReader::NewIterator`. `PartitionedFilterBlockTest` is parameterized to run with different UDT test modes:kNone, kNormal, kStripUserDefinedTimestamp. ``` make all check ./block_based_table_reader_test ./partitioned_filter_block_test ``` Reviewed By: ltamasi Differential Revision: D46344577 Pulled By: jowlyzhang fbshipit-source-id: 93ac8542b19319d1298712b8bed908c8831ba675
2023-06-01 18:10:03 +00:00
bool user_defined_timestamps_persisted() const {
assert(table_ != nullptr);
assert(table_->get_rep() != nullptr);
return table_->get_rep()->user_defined_timestamps_persisted;
}
Use user-provided ReadOptions for metadata block reads more often (#11208) Summary: This is mostly taken from https://github.com/facebook/rocksdb/issues/10427 with my own comments addressed. This PR plumbs the user’s `ReadOptions` down to `GetOrReadIndexBlock()`, `GetOrReadFilterBlock()`, and `GetFilterPartitionBlock()`. Now those functions no longer have to make up a `ReadOptions` with incomplete information. I also let `PartitionIndexReader::NewIterator()` pass through its caller's `ReadOptions::verify_checksums`, which was inexplicably dropped previously. Fixes https://github.com/facebook/rocksdb/issues/10463 Pull Request resolved: https://github.com/facebook/rocksdb/pull/11208 Test Plan: Functional: - Measured `-verify_checksum=false` applies to metadata blocks read outside of table open - setup command: `TEST_TMPDIR=/tmp/100M-DB/ ./db_bench -benchmarks=filluniquerandom,waitforcompaction -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -compression_type=none -num=1638400 -key_size=8 -value_size=56` - run command: `TEST_TMPDIR=/tmp/100M-DB/ ./db_bench -benchmarks=readrandom -use_existing_db=true -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -compression_type=none -num=1638400 -key_size=8 -value_size=56 -duration=10 -threads=32 -cache_size=131072 -statistics=true -verify_checksum=false -open_files=20 -cache_index_and_filter_blocks=true` - before: `rocksdb.block.checksum.compute.count COUNT : 384353` - after: `rocksdb.block.checksum.compute.count COUNT : 22` Performance: - Setup command (tmpfs, 128MB logical data size, cache indexes/filters without pinning so index/filter lookups go through table reader): `TEST_TMPDIR=/dev/shm/128M-DB/ ./db_bench -benchmarks=filluniquerandom,waitforcompaction -write_buffer_size=131072 -target_file_size_base=131072 -max_bytes_for_level_base=524288 -compression_type=none -num=4194304 -key_size=8 -value_size=24 -bloom_bits=8 -whole_key_filtering=1` - Measured point lookup performance. Database is fully cached to emphasize any new callstack overheads - Command: `TEST_TMPDIR=/dev/shm/128M-DB/ ./db_bench -benchmarks=readrandom[-W1][-X20] -use_existing_db=true -cache_index_and_filter_blocks=true -disable_auto_compactions=true -num=4194304 -key_size=8 -value_size=24 -bloom_bits=8 -whole_key_filtering=1 -duration=10 -cache_size=1048576000` - Before: `readrandom [AVG 20 runs] : 274848 (± 3717) ops/sec; 8.4 (± 0.1) MB/sec` - After: `readrandom [AVG 20 runs] : 277904 (± 4474) ops/sec; 8.5 (± 0.1) MB/sec` Reviewed By: hx235 Differential Revision: D43145366 Pulled By: ajkr fbshipit-source-id: 75ec062ece86a82cd788783de9de2c72df57f994
2023-04-04 23:53:14 +00:00
Status GetOrReadIndexBlock(bool no_io, GetContext* get_context,
BlockCacheLookupContext* lookup_context,
Use user-provided ReadOptions for metadata block reads more often (#11208) Summary: This is mostly taken from https://github.com/facebook/rocksdb/issues/10427 with my own comments addressed. This PR plumbs the user’s `ReadOptions` down to `GetOrReadIndexBlock()`, `GetOrReadFilterBlock()`, and `GetFilterPartitionBlock()`. Now those functions no longer have to make up a `ReadOptions` with incomplete information. I also let `PartitionIndexReader::NewIterator()` pass through its caller's `ReadOptions::verify_checksums`, which was inexplicably dropped previously. Fixes https://github.com/facebook/rocksdb/issues/10463 Pull Request resolved: https://github.com/facebook/rocksdb/pull/11208 Test Plan: Functional: - Measured `-verify_checksum=false` applies to metadata blocks read outside of table open - setup command: `TEST_TMPDIR=/tmp/100M-DB/ ./db_bench -benchmarks=filluniquerandom,waitforcompaction -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -compression_type=none -num=1638400 -key_size=8 -value_size=56` - run command: `TEST_TMPDIR=/tmp/100M-DB/ ./db_bench -benchmarks=readrandom -use_existing_db=true -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -compression_type=none -num=1638400 -key_size=8 -value_size=56 -duration=10 -threads=32 -cache_size=131072 -statistics=true -verify_checksum=false -open_files=20 -cache_index_and_filter_blocks=true` - before: `rocksdb.block.checksum.compute.count COUNT : 384353` - after: `rocksdb.block.checksum.compute.count COUNT : 22` Performance: - Setup command (tmpfs, 128MB logical data size, cache indexes/filters without pinning so index/filter lookups go through table reader): `TEST_TMPDIR=/dev/shm/128M-DB/ ./db_bench -benchmarks=filluniquerandom,waitforcompaction -write_buffer_size=131072 -target_file_size_base=131072 -max_bytes_for_level_base=524288 -compression_type=none -num=4194304 -key_size=8 -value_size=24 -bloom_bits=8 -whole_key_filtering=1` - Measured point lookup performance. Database is fully cached to emphasize any new callstack overheads - Command: `TEST_TMPDIR=/dev/shm/128M-DB/ ./db_bench -benchmarks=readrandom[-W1][-X20] -use_existing_db=true -cache_index_and_filter_blocks=true -disable_auto_compactions=true -num=4194304 -key_size=8 -value_size=24 -bloom_bits=8 -whole_key_filtering=1 -duration=10 -cache_size=1048576000` - Before: `readrandom [AVG 20 runs] : 274848 (± 3717) ops/sec; 8.4 (± 0.1) MB/sec` - After: `readrandom [AVG 20 runs] : 277904 (± 4474) ops/sec; 8.5 (± 0.1) MB/sec` Reviewed By: hx235 Differential Revision: D43145366 Pulled By: ajkr fbshipit-source-id: 75ec062ece86a82cd788783de9de2c72df57f994
2023-04-04 23:53:14 +00:00
CachableEntry<Block>* index_block,
const ReadOptions& read_options) const;
size_t ApproximateIndexBlockMemoryUsage() const {
assert(!index_block_.GetOwnValue() || index_block_.GetValue() != nullptr);
return index_block_.GetOwnValue()
? index_block_.GetValue()->ApproximateMemoryUsage()
: 0;
}
private:
const BlockBasedTable* table_;
CachableEntry<Block> index_block_;
};
} // namespace ROCKSDB_NAMESPACE