You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/table/block_based/block_based_table_factory.cc

987 lines
42 KiB

// 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.
#include "table/block_based/block_based_table_factory.h"
#include <stdint.h>
#include <cinttypes>
#include <memory>
#include <string>
#include "cache/cache_entry_roles.h"
#include "logging/logging.h"
Implement XXH3 block checksum type (#9069) Summary: XXH3 - latest hash function that is extremely fast on large data, easily faster than crc32c on most any x86_64 hardware. In integrating this hash function, I have handled the compression type byte in a non-standard way to avoid using the streaming API (extra data movement and active code size because of hash function complexity). This approach got a thumbs-up from Yann Collet. Existing functionality change: * reject bad ChecksumType in options with InvalidArgument This change split off from https://github.com/facebook/rocksdb/issues/9058 because context-aware checksum is likely to be handled through different configuration than ChecksumType. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9069 Test Plan: tests updated, and substantially expanded. Unit tests now check that we don't accidentally change the values generated by the checksum algorithms ("schema test") and that we properly handle invalid/unrecognized checksum types in options or in file footer. DBTestBase::ChangeOptions (etc.) updated from two to one configuration changing from default CRC32c ChecksumType. The point of this test code is to detect possible interactions among features, and the likelihood of some bad interaction being detected by including configurations other than XXH3 and CRC32c--and then not detected by stress/crash test--is extremely low. Stress/crash test also updated (manual run long enough to see it accepts new checksum type). db_bench also updated for microbenchmarking checksums. ### Performance microbenchmark (PORTABLE=0 DEBUG_LEVEL=0, Broadwell processor) ./db_bench -benchmarks=crc32c,xxhash,xxhash64,xxh3,crc32c,xxhash,xxhash64,xxh3,crc32c,xxhash,xxhash64,xxh3 crc32c : 0.200 micros/op 5005220 ops/sec; 19551.6 MB/s (4096 per op) xxhash : 0.807 micros/op 1238408 ops/sec; 4837.5 MB/s (4096 per op) xxhash64 : 0.421 micros/op 2376514 ops/sec; 9283.3 MB/s (4096 per op) xxh3 : 0.171 micros/op 5858391 ops/sec; 22884.3 MB/s (4096 per op) crc32c : 0.206 micros/op 4859566 ops/sec; 18982.7 MB/s (4096 per op) xxhash : 0.793 micros/op 1260850 ops/sec; 4925.2 MB/s (4096 per op) xxhash64 : 0.410 micros/op 2439182 ops/sec; 9528.1 MB/s (4096 per op) xxh3 : 0.161 micros/op 6202872 ops/sec; 24230.0 MB/s (4096 per op) crc32c : 0.203 micros/op 4924686 ops/sec; 19237.1 MB/s (4096 per op) xxhash : 0.839 micros/op 1192388 ops/sec; 4657.8 MB/s (4096 per op) xxhash64 : 0.424 micros/op 2357391 ops/sec; 9208.6 MB/s (4096 per op) xxh3 : 0.162 micros/op 6182678 ops/sec; 24151.1 MB/s (4096 per op) As you can see, especially once warmed up, xxh3 is fastest. ### Performance macrobenchmark (PORTABLE=0 DEBUG_LEVEL=0, Broadwell processor) Test for I in `seq 1 50`; do for CHK in 0 1 2 3 4; do TEST_TMPDIR=/dev/shm/rocksdb$CHK ./db_bench -benchmarks=fillseq -memtablerep=vector -allow_concurrent_memtable_write=false -num=30000000 -checksum_type=$CHK 2>&1 | grep 'micros/op' | tee -a results-$CHK & done; wait; done Results (ops/sec) for FILE in results*; do echo -n "$FILE "; awk '{ s += $5; c++; } END { print 1.0 * s / c; }' < $FILE; done results-0 252118 # kNoChecksum results-1 251588 # kCRC32c results-2 251863 # kxxHash results-3 252016 # kxxHash64 results-4 252038 # kXXH3 Reviewed By: mrambacher Differential Revision: D31905249 Pulled By: pdillinger fbshipit-source-id: cb9b998ebe2523fc7c400eedf62124a78bf4b4d1
3 years ago
#include "options/options_helper.h"
#include "port/port.h"
#include "rocksdb/cache.h"
#include "rocksdb/convenience.h"
#include "rocksdb/filter_policy.h"
#include "rocksdb/flush_block_policy.h"
Implement XXH3 block checksum type (#9069) Summary: XXH3 - latest hash function that is extremely fast on large data, easily faster than crc32c on most any x86_64 hardware. In integrating this hash function, I have handled the compression type byte in a non-standard way to avoid using the streaming API (extra data movement and active code size because of hash function complexity). This approach got a thumbs-up from Yann Collet. Existing functionality change: * reject bad ChecksumType in options with InvalidArgument This change split off from https://github.com/facebook/rocksdb/issues/9058 because context-aware checksum is likely to be handled through different configuration than ChecksumType. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9069 Test Plan: tests updated, and substantially expanded. Unit tests now check that we don't accidentally change the values generated by the checksum algorithms ("schema test") and that we properly handle invalid/unrecognized checksum types in options or in file footer. DBTestBase::ChangeOptions (etc.) updated from two to one configuration changing from default CRC32c ChecksumType. The point of this test code is to detect possible interactions among features, and the likelihood of some bad interaction being detected by including configurations other than XXH3 and CRC32c--and then not detected by stress/crash test--is extremely low. Stress/crash test also updated (manual run long enough to see it accepts new checksum type). db_bench also updated for microbenchmarking checksums. ### Performance microbenchmark (PORTABLE=0 DEBUG_LEVEL=0, Broadwell processor) ./db_bench -benchmarks=crc32c,xxhash,xxhash64,xxh3,crc32c,xxhash,xxhash64,xxh3,crc32c,xxhash,xxhash64,xxh3 crc32c : 0.200 micros/op 5005220 ops/sec; 19551.6 MB/s (4096 per op) xxhash : 0.807 micros/op 1238408 ops/sec; 4837.5 MB/s (4096 per op) xxhash64 : 0.421 micros/op 2376514 ops/sec; 9283.3 MB/s (4096 per op) xxh3 : 0.171 micros/op 5858391 ops/sec; 22884.3 MB/s (4096 per op) crc32c : 0.206 micros/op 4859566 ops/sec; 18982.7 MB/s (4096 per op) xxhash : 0.793 micros/op 1260850 ops/sec; 4925.2 MB/s (4096 per op) xxhash64 : 0.410 micros/op 2439182 ops/sec; 9528.1 MB/s (4096 per op) xxh3 : 0.161 micros/op 6202872 ops/sec; 24230.0 MB/s (4096 per op) crc32c : 0.203 micros/op 4924686 ops/sec; 19237.1 MB/s (4096 per op) xxhash : 0.839 micros/op 1192388 ops/sec; 4657.8 MB/s (4096 per op) xxhash64 : 0.424 micros/op 2357391 ops/sec; 9208.6 MB/s (4096 per op) xxh3 : 0.162 micros/op 6182678 ops/sec; 24151.1 MB/s (4096 per op) As you can see, especially once warmed up, xxh3 is fastest. ### Performance macrobenchmark (PORTABLE=0 DEBUG_LEVEL=0, Broadwell processor) Test for I in `seq 1 50`; do for CHK in 0 1 2 3 4; do TEST_TMPDIR=/dev/shm/rocksdb$CHK ./db_bench -benchmarks=fillseq -memtablerep=vector -allow_concurrent_memtable_write=false -num=30000000 -checksum_type=$CHK 2>&1 | grep 'micros/op' | tee -a results-$CHK & done; wait; done Results (ops/sec) for FILE in results*; do echo -n "$FILE "; awk '{ s += $5; c++; } END { print 1.0 * s / c; }' < $FILE; done results-0 252118 # kNoChecksum results-1 251588 # kCRC32c results-2 251863 # kxxHash results-3 252016 # kxxHash64 results-4 252038 # kXXH3 Reviewed By: mrambacher Differential Revision: D31905249 Pulled By: pdillinger fbshipit-source-id: cb9b998ebe2523fc7c400eedf62124a78bf4b4d1
3 years ago
#include "rocksdb/rocksdb_namespace.h"
#include "rocksdb/table.h"
#include "rocksdb/utilities/options_type.h"
#include "table/block_based/block_based_table_builder.h"
#include "table/block_based/block_based_table_reader.h"
#include "table/format.h"
#include "util/mutexlock.h"
#include "util/string_util.h"
namespace ROCKSDB_NAMESPACE {
void TailPrefetchStats::RecordEffectiveSize(size_t len) {
MutexLock l(&mutex_);
if (num_records_ < kNumTracked) {
num_records_++;
}
records_[next_++] = len;
if (next_ == kNumTracked) {
next_ = 0;
}
}
size_t TailPrefetchStats::GetSuggestedPrefetchSize() {
std::vector<size_t> sorted;
{
MutexLock l(&mutex_);
if (num_records_ == 0) {
return 0;
}
sorted.assign(records_, records_ + num_records_);
}
// Of the historic size, we find the maximum one that satisifis the condtiion
// that if prefetching all, less than 1/8 will be wasted.
std::sort(sorted.begin(), sorted.end());
// Assuming we have 5 data points, and after sorting it looks like this:
//
// +---+
// +---+ | |
// | | | |
// | | | |
// | | | |
// | | | |
// +---+ | | | |
// | | | | | |
// +---+ | | | | | |
// | | | | | | | |
// +---+ | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// +---+ +---+ +---+ +---+ +---+
//
// and we use every of the value as a candidate, and estimate how much we
// wasted, compared to read. For example, when we use the 3rd record
// as candiate. This area is what we read:
// +---+
// +---+ | |
// | | | |
// | | | |
// | | | |
// | | | |
// *** *** *** ***+ *** *** *** *** **
// * | | | | | |
// +---+ | | | | | *
// * | | | | | | | |
// +---+ | | | | | | | *
// * | | | | X | | | | |
// | | | | | | | | | *
// * | | | | | | | | |
// | | | | | | | | | *
// * | | | | | | | | |
// *** *** ***-*** ***--*** ***--*** +****
// which is (size of the record) X (number of records).
//
// While wasted is this area:
// +---+
// +---+ | |
// | | | |
// | | | |
// | | | |
// | | | |
// *** *** *** ****---+ | | | |
// * * | | | | |
// * *-*** *** | | | | |
// * * | | | | | | |
// *--** *** | | | | | | |
// | | | | | X | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// +---+ +---+ +---+ +---+ +---+
//
// Which can be calculated iteratively.
// The difference between wasted using 4st and 3rd record, will
// be following area:
// +---+
// +--+ +-+ ++ +-+ +-+ +---+ | |
// + xxxxxxxxxxxxxxxxxxxxxxxx | | | |
// xxxxxxxxxxxxxxxxxxxxxxxx | | | |
// + xxxxxxxxxxxxxxxxxxxxxxxx | | | |
// | xxxxxxxxxxxxxxxxxxxxxxxx | | | |
// +-+ +-+ +-+ ++ +---+ +--+ | | |
// | | | | | | |
// +---+ ++ | | | | | |
// | | | | | | X | | |
// +---+ ++ | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// | | | | | | | | | |
// +---+ +---+ +---+ +---+ +---+
//
// which will be the size difference between 4st and 3rd record,
// times 3, which is number of records before the 4st.
// Here we assume that all data within the prefetch range will be useful. In
// reality, it may not be the case when a partial block is inside the range,
// or there are data in the middle that is not read. We ignore those cases
// for simplicity.
assert(!sorted.empty());
size_t prev_size = sorted[0];
size_t max_qualified_size = sorted[0];
size_t wasted = 0;
for (size_t i = 1; i < sorted.size(); i++) {
size_t read = sorted[i] * sorted.size();
wasted += (sorted[i] - prev_size) * i;
if (wasted <= read / 8) {
max_qualified_size = sorted[i];
}
prev_size = sorted[i];
}
const size_t kMaxPrefetchSize = 512 * 1024; // Never exceed 512KB
return std::min(kMaxPrefetchSize, max_qualified_size);
}
#ifndef ROCKSDB_LITE
const std::string kOptNameMetadataCacheOpts = "metadata_cache_options";
static std::unordered_map<std::string, PinningTier>
pinning_tier_type_string_map = {
{"kFallback", PinningTier::kFallback},
{"kNone", PinningTier::kNone},
{"kFlushedAndSimilar", PinningTier::kFlushedAndSimilar},
{"kAll", PinningTier::kAll}};
static std::unordered_map<std::string, BlockBasedTableOptions::IndexType>
block_base_table_index_type_string_map = {
{"kBinarySearch", BlockBasedTableOptions::IndexType::kBinarySearch},
{"kHashSearch", BlockBasedTableOptions::IndexType::kHashSearch},
{"kTwoLevelIndexSearch",
BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch},
{"kBinarySearchWithFirstKey",
BlockBasedTableOptions::IndexType::kBinarySearchWithFirstKey}};
static std::unordered_map<std::string,
BlockBasedTableOptions::DataBlockIndexType>
block_base_table_data_block_index_type_string_map = {
{"kDataBlockBinarySearch",
BlockBasedTableOptions::DataBlockIndexType::kDataBlockBinarySearch},
{"kDataBlockBinaryAndHash",
BlockBasedTableOptions::DataBlockIndexType::kDataBlockBinaryAndHash}};
static std::unordered_map<std::string,
BlockBasedTableOptions::IndexShorteningMode>
block_base_table_index_shortening_mode_string_map = {
{"kNoShortening",
BlockBasedTableOptions::IndexShorteningMode::kNoShortening},
{"kShortenSeparators",
BlockBasedTableOptions::IndexShorteningMode::kShortenSeparators},
{"kShortenSeparatorsAndSuccessor",
BlockBasedTableOptions::IndexShorteningMode::
kShortenSeparatorsAndSuccessor}};
static std::unordered_map<std::string, OptionTypeInfo>
metadata_cache_options_type_info = {
{"top_level_index_pinning",
OptionTypeInfo::Enum<PinningTier>(
offsetof(struct MetadataCacheOptions, top_level_index_pinning),
&pinning_tier_type_string_map)},
{"partition_pinning",
OptionTypeInfo::Enum<PinningTier>(
offsetof(struct MetadataCacheOptions, partition_pinning),
&pinning_tier_type_string_map)},
{"unpartitioned_pinning",
OptionTypeInfo::Enum<PinningTier>(
offsetof(struct MetadataCacheOptions, unpartitioned_pinning),
&pinning_tier_type_string_map)}};
static std::unordered_map<std::string,
BlockBasedTableOptions::PrepopulateBlockCache>
block_base_table_prepopulate_block_cache_string_map = {
{"kDisable", BlockBasedTableOptions::PrepopulateBlockCache::kDisable},
{"kFlushOnly",
BlockBasedTableOptions::PrepopulateBlockCache::kFlushOnly}};
#endif // ROCKSDB_LITE
static std::unordered_map<std::string, OptionTypeInfo>
block_based_table_type_info = {
#ifndef ROCKSDB_LITE
/* currently not supported
std::shared_ptr<Cache> block_cache = nullptr;
std::shared_ptr<Cache> block_cache_compressed = nullptr;
*/
{"flush_block_policy_factory",
OptionTypeInfo::AsCustomSharedPtr<FlushBlockPolicyFactory>(
offsetof(struct BlockBasedTableOptions,
flush_block_policy_factory),
OptionVerificationType::kByName, OptionTypeFlags::kCompareNever)},
{"cache_index_and_filter_blocks",
{offsetof(struct BlockBasedTableOptions,
cache_index_and_filter_blocks),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"cache_index_and_filter_blocks_with_high_priority",
{offsetof(struct BlockBasedTableOptions,
cache_index_and_filter_blocks_with_high_priority),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"pin_l0_filter_and_index_blocks_in_cache",
{offsetof(struct BlockBasedTableOptions,
pin_l0_filter_and_index_blocks_in_cache),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"index_type", OptionTypeInfo::Enum<BlockBasedTableOptions::IndexType>(
offsetof(struct BlockBasedTableOptions, index_type),
&block_base_table_index_type_string_map)},
{"hash_index_allow_collision",
{offsetof(struct BlockBasedTableOptions, hash_index_allow_collision),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"data_block_index_type",
OptionTypeInfo::Enum<BlockBasedTableOptions::DataBlockIndexType>(
offsetof(struct BlockBasedTableOptions, data_block_index_type),
&block_base_table_data_block_index_type_string_map)},
{"index_shortening",
OptionTypeInfo::Enum<BlockBasedTableOptions::IndexShorteningMode>(
offsetof(struct BlockBasedTableOptions, index_shortening),
&block_base_table_index_shortening_mode_string_map)},
{"data_block_hash_table_util_ratio",
{offsetof(struct BlockBasedTableOptions,
data_block_hash_table_util_ratio),
OptionType::kDouble, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"checksum",
{offsetof(struct BlockBasedTableOptions, checksum),
OptionType::kChecksumType, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"no_block_cache",
{offsetof(struct BlockBasedTableOptions, no_block_cache),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"block_size",
{offsetof(struct BlockBasedTableOptions, block_size),
OptionType::kSizeT, OptionVerificationType::kNormal,
OptionTypeFlags::kMutable}},
{"block_size_deviation",
{offsetof(struct BlockBasedTableOptions, block_size_deviation),
OptionType::kInt, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"block_restart_interval",
{offsetof(struct BlockBasedTableOptions, block_restart_interval),
OptionType::kInt, OptionVerificationType::kNormal,
OptionTypeFlags::kMutable}},
{"index_block_restart_interval",
{offsetof(struct BlockBasedTableOptions, index_block_restart_interval),
OptionType::kInt, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"index_per_partition",
{0, OptionType::kUInt64T, OptionVerificationType::kDeprecated,
OptionTypeFlags::kNone}},
{"metadata_block_size",
{offsetof(struct BlockBasedTableOptions, metadata_block_size),
OptionType::kUInt64T, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"partition_filters",
{offsetof(struct BlockBasedTableOptions, partition_filters),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
Minimize memory internal fragmentation for Bloom filters (#6427) Summary: New experimental option BBTO::optimize_filters_for_memory builds filters that maximize their use of "usable size" from malloc_usable_size, which is also used to compute block cache charges. Rather than always "rounding up," we track state in the BloomFilterPolicy object to mix essentially "rounding down" and "rounding up" so that the average FP rate of all generated filters is the same as without the option. (YMMV as heavily accessed filters might be unluckily lower accuracy.) Thus, the option near-minimizes what the block cache considers as "memory used" for a given target Bloom filter false positive rate and Bloom filter implementation. There are no forward or backward compatibility issues with this change, though it only works on the format_version=5 Bloom filter. With Jemalloc, we see about 10% reduction in memory footprint (and block cache charge) for Bloom filters, but 1-2% increase in storage footprint, due to encoding efficiency losses (FP rate is non-linear with bits/key). Why not weighted random round up/down rather than state tracking? By only requiring malloc_usable_size, we don't actually know what the next larger and next smaller usable sizes for the allocator are. We pick a requested size, accept and use whatever usable size it has, and use the difference to inform our next choice. This allows us to narrow in on the right balance without tracking/predicting usable sizes. Why not weight history of generated filter false positive rates by number of keys? This could lead to excess skew in small filters after generating a large filter. Results from filter_bench with jemalloc (irrelevant details omitted): (normal keys/filter, but high variance) $ ./filter_bench -quick -impl=2 -average_keys_per_filter=30000 -vary_key_count_ratio=0.9 Build avg ns/key: 29.6278 Number of filters: 5516 Total size (MB): 200.046 Reported total allocated memory (MB): 220.597 Reported internal fragmentation: 10.2732% Bits/key stored: 10.0097 Average FP rate %: 0.965228 $ ./filter_bench -quick -impl=2 -average_keys_per_filter=30000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory Build avg ns/key: 30.5104 Number of filters: 5464 Total size (MB): 200.015 Reported total allocated memory (MB): 200.322 Reported internal fragmentation: 0.153709% Bits/key stored: 10.1011 Average FP rate %: 0.966313 (very few keys / filter, optimization not as effective due to ~59 byte internal fragmentation in blocked Bloom filter representation) $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000 -vary_key_count_ratio=0.9 Build avg ns/key: 29.5649 Number of filters: 162950 Total size (MB): 200.001 Reported total allocated memory (MB): 224.624 Reported internal fragmentation: 12.3117% Bits/key stored: 10.2951 Average FP rate %: 0.821534 $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory Build avg ns/key: 31.8057 Number of filters: 159849 Total size (MB): 200 Reported total allocated memory (MB): 208.846 Reported internal fragmentation: 4.42297% Bits/key stored: 10.4948 Average FP rate %: 0.811006 (high keys/filter) $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000000 -vary_key_count_ratio=0.9 Build avg ns/key: 29.7017 Number of filters: 164 Total size (MB): 200.352 Reported total allocated memory (MB): 221.5 Reported internal fragmentation: 10.5552% Bits/key stored: 10.0003 Average FP rate %: 0.969358 $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory Build avg ns/key: 30.7131 Number of filters: 160 Total size (MB): 200.928 Reported total allocated memory (MB): 200.938 Reported internal fragmentation: 0.00448054% Bits/key stored: 10.1852 Average FP rate %: 0.963387 And from db_bench (block cache) with jemalloc: $ ./db_bench -db=/dev/shm/dbbench.no_optimize -benchmarks=fillrandom -format_version=5 -value_size=90 -bloom_bits=10 -num=2000000 -threads=8 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false $ ./db_bench -db=/dev/shm/dbbench -benchmarks=fillrandom -format_version=5 -value_size=90 -bloom_bits=10 -num=2000000 -threads=8 -optimize_filters_for_memory -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false $ (for FILE in /dev/shm/dbbench.no_optimize/*.sst; do ./sst_dump --file=$FILE --show_properties | grep 'filter block' ; done) | awk '{ t += $4; } END { print t; }' 17063835 $ (for FILE in /dev/shm/dbbench/*.sst; do ./sst_dump --file=$FILE --show_properties | grep 'filter block' ; done) | awk '{ t += $4; } END { print t; }' 17430747 $ #^ 2.1% additional filter storage $ ./db_bench -db=/dev/shm/dbbench.no_optimize -use_existing_db -benchmarks=readrandom,stats -statistics -bloom_bits=10 -num=2000000 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false -duration=10 -cache_index_and_filter_blocks -cache_size=1000000000 rocksdb.block.cache.index.add COUNT : 33 rocksdb.block.cache.index.bytes.insert COUNT : 8440400 rocksdb.block.cache.filter.add COUNT : 33 rocksdb.block.cache.filter.bytes.insert COUNT : 21087528 rocksdb.bloom.filter.useful COUNT : 4963889 rocksdb.bloom.filter.full.positive COUNT : 1214081 rocksdb.bloom.filter.full.true.positive COUNT : 1161999 $ #^ 1.04 % observed FP rate $ ./db_bench -db=/dev/shm/dbbench -use_existing_db -benchmarks=readrandom,stats -statistics -bloom_bits=10 -num=2000000 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false -optimize_filters_for_memory -duration=10 -cache_index_and_filter_blocks -cache_size=1000000000 rocksdb.block.cache.index.add COUNT : 33 rocksdb.block.cache.index.bytes.insert COUNT : 8448592 rocksdb.block.cache.filter.add COUNT : 33 rocksdb.block.cache.filter.bytes.insert COUNT : 18220328 rocksdb.bloom.filter.useful COUNT : 5360933 rocksdb.bloom.filter.full.positive COUNT : 1321315 rocksdb.bloom.filter.full.true.positive COUNT : 1262999 $ #^ 1.08 % observed FP rate, 13.6% less memory usage for filters (Due to specific key density, this example tends to generate filters that are "worse than average" for internal fragmentation. "Better than average" cases can show little or no improvement.) Pull Request resolved: https://github.com/facebook/rocksdb/pull/6427 Test Plan: unit test added, 'make check' with gcc, clang and valgrind Reviewed By: siying Differential Revision: D22124374 Pulled By: pdillinger fbshipit-source-id: f3e3aa152f9043ddf4fae25799e76341d0d8714e
4 years ago
{"optimize_filters_for_memory",
{offsetof(struct BlockBasedTableOptions, optimize_filters_for_memory),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"filter_policy",
{offsetof(struct BlockBasedTableOptions, filter_policy),
OptionType::kUnknown, OptionVerificationType::kByNameAllowFromNull,
OptionTypeFlags::kNone,
// Parses the Filter policy
[](const ConfigOptions& opts, const std::string&,
const std::string& value, void* addr) {
auto* policy =
static_cast<std::shared_ptr<const FilterPolicy>*>(addr);
return FilterPolicy::CreateFromString(opts, value, policy);
},
// Converts the FilterPolicy to its string representation
[](const ConfigOptions&, const std::string&, const void* addr,
std::string* value) {
const auto* policy =
static_cast<const std::shared_ptr<const FilterPolicy>*>(addr);
if (policy->get()) {
*value = (*policy)->Name();
} else {
*value = kNullptrString;
}
return Status::OK();
},
// Compares two FilterPolicy objects for equality
[](const ConfigOptions&, const std::string&, const void* addr1,
const void* addr2, std::string*) {
const auto* policy1 =
static_cast<const std::shared_ptr<const FilterPolicy>*>(addr1)
->get();
const auto* policy2 =
static_cast<const std::shared_ptr<FilterPolicy>*>(addr2)->get();
if (policy1 == policy2) {
return true;
} else if (policy1 != nullptr && policy2 != nullptr) {
return (strcmp(policy1->Name(), policy2->Name()) == 0);
} else {
return false;
}
}}},
{"whole_key_filtering",
{offsetof(struct BlockBasedTableOptions, whole_key_filtering),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
Detect (new) Bloom/Ribbon Filter construction corruption (#9342) Summary: Note: rebase on and merge after https://github.com/facebook/rocksdb/pull/9349, https://github.com/facebook/rocksdb/pull/9345, (optional) https://github.com/facebook/rocksdb/pull/9393 **Context:** (Quoted from pdillinger) Layers of information during new Bloom/Ribbon Filter construction in building block-based tables includes the following: a) set of keys to add to filter b) set of hashes to add to filter (64-bit hash applied to each key) c) set of Bloom indices to set in filter, with duplicates d) set of Bloom indices to set in filter, deduplicated e) final filter and its checksum This PR aims to detect corruption (e.g, unexpected hardware/software corruption on data structures residing in the memory for a long time) from b) to e) and leave a) as future works for application level. - b)'s corruption is detected by verifying the xor checksum of the hash entries calculated as the entries accumulate before being added to the filter. (i.e, `XXPH3FilterBitsBuilder::MaybeVerifyHashEntriesChecksum()`) - c) - e)'s corruption is detected by verifying the hash entries indeed exists in the constructed filter by re-querying these hash entries in the filter (i.e, `FilterBitsBuilder::MaybePostVerify()`) after computing the block checksum (except for PartitionFilter, which is done right after each `FilterBitsBuilder::Finish` for impl simplicity - see code comment for more). For this stage of detection, we assume hash entries are not corrupted after checking on b) since the time interval from b) to c) is relatively short IMO. Option to enable this feature of detection is `BlockBasedTableOptions::detect_filter_construct_corruption` which is false by default. **Summary:** - Implemented new functions `XXPH3FilterBitsBuilder::MaybeVerifyHashEntriesChecksum()` and `FilterBitsBuilder::MaybePostVerify()` - Ensured hash entries, final filter and banding and their [cache reservation ](https://github.com/facebook/rocksdb/issues/9073) are released properly despite corruption - See [Filter.construction.artifacts.release.point.pdf ](https://github.com/facebook/rocksdb/files/7923487/Design.Filter.construction.artifacts.release.point.pdf) for high-level design - Bundled and refactored hash entries's related artifact in XXPH3FilterBitsBuilder into `HashEntriesInfo` for better control on lifetime of these artifact during `SwapEntires`, `ResetEntries` - Ensured RocksDB block-based table builder calls `FilterBitsBuilder::MaybePostVerify()` after constructing the filter by `FilterBitsBuilder::Finish()` - When encountering such filter construction corruption, stop writing the filter content to files and mark such a block-based table building non-ok by storing the corruption status in the builder. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9342 Test Plan: - Added new unit test `DBFilterConstructionCorruptionTestWithParam.DetectCorruption` - Included this new feature in `DBFilterConstructionReserveMemoryTestWithParam.ReserveMemory` as this feature heavily touch ReserveMemory's impl - For fallback case, I run `./filter_bench -impl=3 -detect_filter_construct_corruption=true -reserve_table_builder_memory=true -strict_capacity_limit=true -quick -runs 10 | grep 'Build avg'` to make sure nothing break. - Added to `filter_bench`: increased filter construction time by **30%**, mostly by `MaybePostVerify()` - FastLocalBloom - Before change: `./filter_bench -impl=2 -quick -runs 10 | grep 'Build avg'`: **28.86643s** - After change: - `./filter_bench -impl=2 -detect_filter_construct_corruption=false -quick -runs 10 | grep 'Build avg'` (expect a tiny increase due to MaybePostVerify is always called regardless): **27.6644s (-4% perf improvement might be due to now we don't drop bloom hash entry in `AddAllEntries` along iteration but in bulk later, same with the bypassing-MaybePostVerify case below)** - `./filter_bench -impl=2 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'` (expect acceptable increase): **34.41159s (+20%)** - `./filter_bench -impl=2 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'` (by-passing MaybePostVerify, expect minor increase): **27.13431s (-6%)** - Standard128Ribbon - Before change: `./filter_bench -impl=3 -quick -runs 10 | grep 'Build avg'`: **122.5384s** - After change: - `./filter_bench -impl=3 -detect_filter_construct_corruption=false -quick -runs 10 | grep 'Build avg'` (expect a tiny increase due to MaybePostVerify is always called regardless - verified by removing MaybePostVerify under this case and found only +-1ns difference): **124.3588s (+2%)** - `./filter_bench -impl=3 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'`(expect acceptable increase): **159.4946s (+30%)** - `./filter_bench -impl=3 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'`(by-passing MaybePostVerify, expect minor increase) : **125.258s (+2%)** - Added to `db_stress`: `make crash_test`, `./db_stress --detect_filter_construct_corruption=true` - Manually smoke-tested: manually corrupted the filter construction in some db level tests with basic PUT and background flush. As expected, the error did get returned to users in subsequent PUT and Flush status. Reviewed By: pdillinger Differential Revision: D33746928 Pulled By: hx235 fbshipit-source-id: cb056426be5a7debc1cd16f23bc250f36a08ca57
3 years ago
{"detect_filter_construct_corruption",
{offsetof(struct BlockBasedTableOptions,
detect_filter_construct_corruption),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
Account Bloom/Ribbon filter construction memory in global memory limit (#9073) Summary: Note: This PR is the 4th part of a bigger PR stack (https://github.com/facebook/rocksdb/pull/9073) and will rebase/merge only after the first three PRs (https://github.com/facebook/rocksdb/pull/9070, https://github.com/facebook/rocksdb/pull/9071, https://github.com/facebook/rocksdb/pull/9130) merge. **Context:** Similar to https://github.com/facebook/rocksdb/pull/8428, this PR is to track memory usage during (new) Bloom Filter (i.e,FastLocalBloom) and Ribbon Filter (i.e, Ribbon128) construction, moving toward the goal of [single global memory limit using block cache capacity](https://github.com/facebook/rocksdb/wiki/Projects-Being-Developed#improving-memory-efficiency). It also constrains the size of the banding portion of Ribbon Filter during construction by falling back to Bloom Filter if that banding is, at some point, larger than the available space in the cache under `LRUCacheOptions::strict_capacity_limit=true`. The option to turn on this feature is `BlockBasedTableOptions::reserve_table_builder_memory = true` which by default is set to `false`. We [decided](https://github.com/facebook/rocksdb/pull/9073#discussion_r741548409) not to have separate option for separate memory user in table building therefore their memory accounting are all bundled under one general option. **Summary:** - Reserved/released cache for creation/destruction of three main memory users with the passed-in `FilterBuildingContext::cache_res_mgr` during filter construction: - hash entries (i.e`hash_entries`.size(), we bucket-charge hash entries during insertion for performance), - banding (Ribbon Filter only, `bytes_coeff_rows` +`bytes_result_rows` + `bytes_backtrack`), - final filter (i.e, `mutable_buf`'s size). - Implementation details: in order to use `CacheReservationManager::CacheReservationHandle` to account final filter's memory, we have to store the `CacheReservationManager` object and `CacheReservationHandle` for final filter in `XXPH3BitsFilterBuilder` as well as explicitly delete the filter bits builder when done with the final filter in block based table. - Added option fo run `filter_bench` with this memory reservation feature Pull Request resolved: https://github.com/facebook/rocksdb/pull/9073 Test Plan: - Added new tests in `db_bloom_filter_test` to verify filter construction peak cache reservation under combination of `BlockBasedTable::Rep::FilterType` (e.g, `kFullFilter`, `kPartitionedFilter`), `BloomFilterPolicy::Mode`(e.g, `kFastLocalBloom`, `kStandard128Ribbon`, `kDeprecatedBlock`) and `BlockBasedTableOptions::reserve_table_builder_memory` - To address the concern for slow test: tests with memory reservation under `kFullFilter` + `kStandard128Ribbon` and `kPartitionedFilter` take around **3000 - 6000 ms** and others take around **1500 - 2000 ms**, in total adding **20000 - 25000 ms** to the test suit running locally - Added new test in `bloom_test` to verify Ribbon Filter fallback on large banding in FullFilter - Added test in `filter_bench` to verify that this feature does not significantly slow down Bloom/Ribbon Filter construction speed. Local result averaged over **20** run as below: - FastLocalBloom - baseline `./filter_bench -impl=2 -quick -runs 20 | grep 'Build avg'`: - **Build avg ns/key: 29.56295** (DEBUG_LEVEL=1), **29.98153** (DEBUG_LEVEL=0) - new feature (expected to be similar as above)`./filter_bench -impl=2 -quick -runs 20 -reserve_table_builder_memory=true | grep 'Build avg'`: - **Build avg ns/key: 30.99046** (DEBUG_LEVEL=1), **30.48867** (DEBUG_LEVEL=0) - new feature of RibbonFilter with fallback (expected to be similar as above) `./filter_bench -impl=2 -quick -runs 20 -reserve_table_builder_memory=true -strict_capacity_limit=true | grep 'Build avg'` : - **Build avg ns/key: 31.146975** (DEBUG_LEVEL=1), **30.08165** (DEBUG_LEVEL=0) - Ribbon128 - baseline `./filter_bench -impl=3 -quick -runs 20 | grep 'Build avg'`: - **Build avg ns/key: 129.17585** (DEBUG_LEVEL=1), **130.5225** (DEBUG_LEVEL=0) - new feature (expected to be similar as above) `./filter_bench -impl=3 -quick -runs 20 -reserve_table_builder_memory=true | grep 'Build avg' `: - **Build avg ns/key: 131.61645** (DEBUG_LEVEL=1), **132.98075** (DEBUG_LEVEL=0) - new feature of RibbonFilter with fallback (expected to be a lot faster than above due to fallback) `./filter_bench -impl=3 -quick -runs 20 -reserve_table_builder_memory=true -strict_capacity_limit=true | grep 'Build avg'` : - **Build avg ns/key: 52.032965** (DEBUG_LEVEL=1), **52.597825** (DEBUG_LEVEL=0) - And the warning message of `"Cache reservation for Ribbon filter banding failed due to cache full"` is indeed logged to console. Reviewed By: pdillinger Differential Revision: D31991348 Pulled By: hx235 fbshipit-source-id: 9336b2c60f44d530063da518ceaf56dac5f9df8e
3 years ago
{"reserve_table_builder_memory",
{offsetof(struct BlockBasedTableOptions, reserve_table_builder_memory),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"skip_table_builder_flush",
{0, OptionType::kBoolean, OptionVerificationType::kDeprecated,
OptionTypeFlags::kNone}},
{"format_version",
{offsetof(struct BlockBasedTableOptions, format_version),
OptionType::kUInt32T, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"verify_compression",
{offsetof(struct BlockBasedTableOptions, verify_compression),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"read_amp_bytes_per_bit",
{offsetof(struct BlockBasedTableOptions, read_amp_bytes_per_bit),
OptionType::kUInt32T, OptionVerificationType::kNormal,
OptionTypeFlags::kNone,
[](const ConfigOptions& /*opts*/, const std::string& /*name*/,
const std::string& value, void* addr) {
// A workaround to fix a bug in 6.10, 6.11, 6.12, 6.13
// and 6.14. The bug will write out 8 bytes to OPTIONS file from the
// starting address of BlockBasedTableOptions.read_amp_bytes_per_bit
// which is actually a uint32. Consequently, the value of
// read_amp_bytes_per_bit written in the OPTIONS file is wrong.
// From 6.15, RocksDB will try to parse the read_amp_bytes_per_bit
// from OPTIONS file as a uint32. To be able to load OPTIONS file
// generated by affected releases before the fix, we need to
// manually parse read_amp_bytes_per_bit with this special hack.
uint64_t read_amp_bytes_per_bit = ParseUint64(value);
*(static_cast<uint32_t*>(addr)) =
static_cast<uint32_t>(read_amp_bytes_per_bit);
return Status::OK();
}}},
{"enable_index_compression",
{offsetof(struct BlockBasedTableOptions, enable_index_compression),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"block_align",
{offsetof(struct BlockBasedTableOptions, block_align),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{"pin_top_level_index_and_filter",
{offsetof(struct BlockBasedTableOptions,
pin_top_level_index_and_filter),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kNone}},
{kOptNameMetadataCacheOpts,
OptionTypeInfo::Struct(
kOptNameMetadataCacheOpts, &metadata_cache_options_type_info,
offsetof(struct BlockBasedTableOptions, metadata_cache_options),
OptionVerificationType::kNormal, OptionTypeFlags::kNone)},
{"block_cache",
{offsetof(struct BlockBasedTableOptions, block_cache),
OptionType::kUnknown, OptionVerificationType::kNormal,
(OptionTypeFlags::kCompareNever | OptionTypeFlags::kDontSerialize),
// Parses the input vsalue as a Cache
[](const ConfigOptions& opts, const std::string&,
const std::string& value, void* addr) {
auto* cache = static_cast<std::shared_ptr<Cache>*>(addr);
return Cache::CreateFromString(opts, value, cache);
}}},
{"block_cache_compressed",
{offsetof(struct BlockBasedTableOptions, block_cache_compressed),
OptionType::kUnknown, OptionVerificationType::kNormal,
(OptionTypeFlags::kCompareNever | OptionTypeFlags::kDontSerialize),
// Parses the input vsalue as a Cache
[](const ConfigOptions& opts, const std::string&,
const std::string& value, void* addr) {
auto* cache = static_cast<std::shared_ptr<Cache>*>(addr);
return Cache::CreateFromString(opts, value, cache);
}}},
{"max_auto_readahead_size",
{offsetof(struct BlockBasedTableOptions, max_auto_readahead_size),
OptionType::kSizeT, OptionVerificationType::kNormal,
OptionTypeFlags::kMutable}},
{"prepopulate_block_cache",
OptionTypeInfo::Enum<BlockBasedTableOptions::PrepopulateBlockCache>(
offsetof(struct BlockBasedTableOptions, prepopulate_block_cache),
&block_base_table_prepopulate_block_cache_string_map,
OptionTypeFlags::kMutable)},
#endif // ROCKSDB_LITE
};
// TODO(myabandeh): We should return an error instead of silently changing the
// options
BlockBasedTableFactory::BlockBasedTableFactory(
const BlockBasedTableOptions& _table_options)
: table_options_(_table_options) {
InitializeOptions();
RegisterOptions(&table_options_, &block_based_table_type_info);
}
void BlockBasedTableFactory::InitializeOptions() {
if (table_options_.flush_block_policy_factory == nullptr) {
table_options_.flush_block_policy_factory.reset(
new FlushBlockBySizePolicyFactory());
}
if (table_options_.no_block_cache) {
table_options_.block_cache.reset();
} else if (table_options_.block_cache == nullptr) {
LRUCacheOptions co;
co.capacity = 8 << 20;
// It makes little sense to pay overhead for mid-point insertion while the
// block size is only 8MB.
co.high_pri_pool_ratio = 0.0;
table_options_.block_cache = NewLRUCache(co);
}
if (table_options_.block_size_deviation < 0 ||
table_options_.block_size_deviation > 100) {
table_options_.block_size_deviation = 0;
}
if (table_options_.block_restart_interval < 1) {
table_options_.block_restart_interval = 1;
}
if (table_options_.index_block_restart_interval < 1) {
table_options_.index_block_restart_interval = 1;
}
if (table_options_.index_type == BlockBasedTableOptions::kHashSearch &&
table_options_.index_block_restart_interval != 1) {
// Currently kHashSearch is incompatible with index_block_restart_interval > 1
table_options_.index_block_restart_interval = 1;
}
if (table_options_.partition_filters &&
table_options_.index_type !=
BlockBasedTableOptions::kTwoLevelIndexSearch) {
// We do not support partitioned filters without partitioning indexes
table_options_.partition_filters = false;
}
}
Status BlockBasedTableFactory::PrepareOptions(const ConfigOptions& opts) {
InitializeOptions();
return TableFactory::PrepareOptions(opts);
}
namespace {
// Different cache kinds use the same keys for physically different values, so
// they must not share an underlying key space with each other.
Status CheckCacheOptionCompatibility(const BlockBasedTableOptions& bbto) {
int cache_count = (bbto.block_cache != nullptr) +
(bbto.block_cache_compressed != nullptr) +
(bbto.persistent_cache != nullptr);
if (cache_count <= 1) {
// Nothing to share / overlap
return Status::OK();
}
// Simple pointer equality
if (bbto.block_cache == bbto.block_cache_compressed) {
return Status::InvalidArgument(
"block_cache same as block_cache_compressed not currently supported, "
"and would be bad for performance anyway");
}
// More complex test of shared key space, in case the instances are wrappers
// for some shared underlying cache.
std::string sentinel_key(size_t{1}, '\0');
static char kRegularBlockCacheMarker = 'b';
static char kCompressedBlockCacheMarker = 'c';
static char kPersistentCacheMarker = 'p';
if (bbto.block_cache) {
bbto.block_cache
->Insert(Slice(sentinel_key), &kRegularBlockCacheMarker, 1,
GetNoopDeleterForRole<CacheEntryRole::kMisc>())
.PermitUncheckedError();
}
if (bbto.block_cache_compressed) {
bbto.block_cache_compressed
->Insert(Slice(sentinel_key), &kCompressedBlockCacheMarker, 1,
GetNoopDeleterForRole<CacheEntryRole::kMisc>())
.PermitUncheckedError();
}
if (bbto.persistent_cache) {
// Note: persistent cache copies the data, not keeping the pointer
bbto.persistent_cache
->Insert(Slice(sentinel_key), &kPersistentCacheMarker, 1)
.PermitUncheckedError();
}
// If we get something different from what we inserted, that indicates
// dangerously overlapping key spaces.
if (bbto.block_cache) {
auto handle = bbto.block_cache->Lookup(Slice(sentinel_key));
if (handle) {
auto v = static_cast<char*>(bbto.block_cache->Value(handle));
char c = *v;
bbto.block_cache->Release(handle);
if (v == &kCompressedBlockCacheMarker) {
return Status::InvalidArgument(
"block_cache and block_cache_compressed share the same key space, "
"which is not supported");
} else if (c == kPersistentCacheMarker) {
return Status::InvalidArgument(
"block_cache and persistent_cache share the same key space, "
"which is not supported");
} else if (v != &kRegularBlockCacheMarker) {
return Status::Corruption("Unexpected mutation to block_cache");
}
}
}
if (bbto.block_cache_compressed) {
auto handle = bbto.block_cache_compressed->Lookup(Slice(sentinel_key));
if (handle) {
auto v = static_cast<char*>(bbto.block_cache_compressed->Value(handle));
char c = *v;
bbto.block_cache_compressed->Release(handle);
if (v == &kRegularBlockCacheMarker) {
return Status::InvalidArgument(
"block_cache_compressed and block_cache share the same key space, "
"which is not supported");
} else if (c == kPersistentCacheMarker) {
return Status::InvalidArgument(
"block_cache_compressed and persistent_cache share the same key "
"space, "
"which is not supported");
} else if (v != &kCompressedBlockCacheMarker) {
return Status::Corruption(
"Unexpected mutation to block_cache_compressed");
}
}
}
if (bbto.persistent_cache) {
std::unique_ptr<char[]> data;
size_t size = 0;
bbto.persistent_cache->Lookup(Slice(sentinel_key), &data, &size)
.PermitUncheckedError();
if (data && size > 0) {
if (data[0] == kRegularBlockCacheMarker) {
return Status::InvalidArgument(
"persistent_cache and block_cache share the same key space, "
"which is not supported");
} else if (data[0] == kCompressedBlockCacheMarker) {
return Status::InvalidArgument(
"persistent_cache and block_cache_compressed share the same key "
"space, "
"which is not supported");
} else if (data[0] != kPersistentCacheMarker) {
return Status::Corruption("Unexpected mutation to persistent_cache");
}
}
}
return Status::OK();
}
} // namespace
Status BlockBasedTableFactory::NewTableReader(
const ReadOptions& ro, const TableReaderOptions& table_reader_options,
std::unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
std::unique_ptr<TableReader>* table_reader,
bool prefetch_index_and_filter_in_cache) const {
return BlockBasedTable::Open(
ro, table_reader_options.ioptions, table_reader_options.env_options,
table_options_, table_reader_options.internal_comparator, std::move(file),
file_size, table_reader, table_reader_options.prefix_extractor,
prefetch_index_and_filter_in_cache, table_reader_options.skip_filters,
table_reader_options.level, table_reader_options.immortal,
table_reader_options.largest_seqno,
table_reader_options.force_direct_prefetch, &tail_prefetch_stats_,
table_reader_options.block_cache_tracer,
table_reader_options.max_file_size_for_l0_meta_pin,
table_reader_options.cur_db_session_id,
table_reader_options.cur_file_num);
}
TableBuilder* BlockBasedTableFactory::NewTableBuilder(
const TableBuilderOptions& table_builder_options,
WritableFileWriter* file) const {
return new BlockBasedTableBuilder(table_options_, table_builder_options,
file);
}
Status BlockBasedTableFactory::ValidateOptions(
Unordered Writes (#5218) Summary: Performing unordered writes in rocksdb when unordered_write option is set to true. When enabled the writes to memtable are done without joining any write thread. This offers much higher write throughput since the upcoming writes would not have to wait for the slowest memtable write to finish. The tradeoff is that the writes visible to a snapshot might change over time. If the application cannot tolerate that, it should implement its own mechanisms to work around that. Using TransactionDB with WRITE_PREPARED write policy is one way to achieve that. Doing so increases the max throughput by 2.2x without however compromising the snapshot guarantees. The patch is prepared based on an original by siying Existing unit tests are extended to include unordered_write option. Benchmark Results: ``` TEST_TMPDIR=/dev/shm/ ./db_bench_unordered --benchmarks=fillrandom --threads=32 --num=10000000 -max_write_buffer_number=16 --max_background_jobs=64 --batch_size=8 --writes=3000000 -level0_file_num_compaction_trigger=99999 --level0_slowdown_writes_trigger=99999 --level0_stop_writes_trigger=99999 -enable_pipelined_write=false -disable_auto_compactions --unordered_write=1 ``` With WAL - Vanilla RocksDB: 78.6 MB/s - WRITER_PREPARED with unordered_write: 177.8 MB/s (2.2x) - unordered_write: 368.9 MB/s (4.7x with relaxed snapshot guarantees) Without WAL - Vanilla RocksDB: 111.3 MB/s - WRITER_PREPARED with unordered_write: 259.3 MB/s MB/s (2.3x) - unordered_write: 645.6 MB/s (5.8x with relaxed snapshot guarantees) - WRITER_PREPARED with unordered_write disable concurrency control: 185.3 MB/s MB/s (2.35x) Limitations: - The feature is not yet extended to `max_successive_merges` > 0. The feature is also incompatible with `enable_pipelined_write` = true as well as with `allow_concurrent_memtable_write` = false. Pull Request resolved: https://github.com/facebook/rocksdb/pull/5218 Differential Revision: D15219029 Pulled By: maysamyabandeh fbshipit-source-id: 38f2abc4af8780148c6128acdba2b3227bc81759
6 years ago
const DBOptions& db_opts, const ColumnFamilyOptions& cf_opts) const {
if (table_options_.index_type == BlockBasedTableOptions::kHashSearch &&
cf_opts.prefix_extractor == nullptr) {
return Status::InvalidArgument(
"Hash index is specified for block-based "
"table, but prefix_extractor is not given");
}
if (table_options_.cache_index_and_filter_blocks &&
table_options_.no_block_cache) {
return Status::InvalidArgument(
"Enable cache_index_and_filter_blocks, "
", but block cache is disabled");
}
if (table_options_.pin_l0_filter_and_index_blocks_in_cache &&
table_options_.no_block_cache) {
return Status::InvalidArgument(
"Enable pin_l0_filter_and_index_blocks_in_cache, "
", but block cache is disabled");
}
if (!IsSupportedFormatVersion(table_options_.format_version)) {
return Status::InvalidArgument(
"Unsupported BlockBasedTable format_version. Please check "
"include/rocksdb/table.h for more info");
}
if (table_options_.block_align && (cf_opts.compression != kNoCompression)) {
return Status::InvalidArgument(
"Enable block_align, but compression "
"enabled");
}
if (table_options_.block_align &&
(table_options_.block_size & (table_options_.block_size - 1))) {
return Status::InvalidArgument(
"Block alignment requested but block size is not a power of 2");
}
if (table_options_.block_size > port::kMaxUint32) {
return Status::InvalidArgument(
"block size exceeds maximum number (4GiB) allowed");
}
if (table_options_.data_block_index_type ==
BlockBasedTableOptions::kDataBlockBinaryAndHash &&
table_options_.data_block_hash_table_util_ratio <= 0) {
return Status::InvalidArgument(
"data_block_hash_table_util_ratio should be greater than 0 when "
"data_block_index_type is set to kDataBlockBinaryAndHash");
}
Unordered Writes (#5218) Summary: Performing unordered writes in rocksdb when unordered_write option is set to true. When enabled the writes to memtable are done without joining any write thread. This offers much higher write throughput since the upcoming writes would not have to wait for the slowest memtable write to finish. The tradeoff is that the writes visible to a snapshot might change over time. If the application cannot tolerate that, it should implement its own mechanisms to work around that. Using TransactionDB with WRITE_PREPARED write policy is one way to achieve that. Doing so increases the max throughput by 2.2x without however compromising the snapshot guarantees. The patch is prepared based on an original by siying Existing unit tests are extended to include unordered_write option. Benchmark Results: ``` TEST_TMPDIR=/dev/shm/ ./db_bench_unordered --benchmarks=fillrandom --threads=32 --num=10000000 -max_write_buffer_number=16 --max_background_jobs=64 --batch_size=8 --writes=3000000 -level0_file_num_compaction_trigger=99999 --level0_slowdown_writes_trigger=99999 --level0_stop_writes_trigger=99999 -enable_pipelined_write=false -disable_auto_compactions --unordered_write=1 ``` With WAL - Vanilla RocksDB: 78.6 MB/s - WRITER_PREPARED with unordered_write: 177.8 MB/s (2.2x) - unordered_write: 368.9 MB/s (4.7x with relaxed snapshot guarantees) Without WAL - Vanilla RocksDB: 111.3 MB/s - WRITER_PREPARED with unordered_write: 259.3 MB/s MB/s (2.3x) - unordered_write: 645.6 MB/s (5.8x with relaxed snapshot guarantees) - WRITER_PREPARED with unordered_write disable concurrency control: 185.3 MB/s MB/s (2.35x) Limitations: - The feature is not yet extended to `max_successive_merges` > 0. The feature is also incompatible with `enable_pipelined_write` = true as well as with `allow_concurrent_memtable_write` = false. Pull Request resolved: https://github.com/facebook/rocksdb/pull/5218 Differential Revision: D15219029 Pulled By: maysamyabandeh fbshipit-source-id: 38f2abc4af8780148c6128acdba2b3227bc81759
6 years ago
if (db_opts.unordered_write && cf_opts.max_successive_merges > 0) {
// TODO(myabandeh): support it
return Status::InvalidArgument(
"max_successive_merges larger than 0 is currently inconsistent with "
"unordered_write");
}
{
Status s = CheckCacheOptionCompatibility(table_options_);
if (!s.ok()) {
return s;
}
}
Implement XXH3 block checksum type (#9069) Summary: XXH3 - latest hash function that is extremely fast on large data, easily faster than crc32c on most any x86_64 hardware. In integrating this hash function, I have handled the compression type byte in a non-standard way to avoid using the streaming API (extra data movement and active code size because of hash function complexity). This approach got a thumbs-up from Yann Collet. Existing functionality change: * reject bad ChecksumType in options with InvalidArgument This change split off from https://github.com/facebook/rocksdb/issues/9058 because context-aware checksum is likely to be handled through different configuration than ChecksumType. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9069 Test Plan: tests updated, and substantially expanded. Unit tests now check that we don't accidentally change the values generated by the checksum algorithms ("schema test") and that we properly handle invalid/unrecognized checksum types in options or in file footer. DBTestBase::ChangeOptions (etc.) updated from two to one configuration changing from default CRC32c ChecksumType. The point of this test code is to detect possible interactions among features, and the likelihood of some bad interaction being detected by including configurations other than XXH3 and CRC32c--and then not detected by stress/crash test--is extremely low. Stress/crash test also updated (manual run long enough to see it accepts new checksum type). db_bench also updated for microbenchmarking checksums. ### Performance microbenchmark (PORTABLE=0 DEBUG_LEVEL=0, Broadwell processor) ./db_bench -benchmarks=crc32c,xxhash,xxhash64,xxh3,crc32c,xxhash,xxhash64,xxh3,crc32c,xxhash,xxhash64,xxh3 crc32c : 0.200 micros/op 5005220 ops/sec; 19551.6 MB/s (4096 per op) xxhash : 0.807 micros/op 1238408 ops/sec; 4837.5 MB/s (4096 per op) xxhash64 : 0.421 micros/op 2376514 ops/sec; 9283.3 MB/s (4096 per op) xxh3 : 0.171 micros/op 5858391 ops/sec; 22884.3 MB/s (4096 per op) crc32c : 0.206 micros/op 4859566 ops/sec; 18982.7 MB/s (4096 per op) xxhash : 0.793 micros/op 1260850 ops/sec; 4925.2 MB/s (4096 per op) xxhash64 : 0.410 micros/op 2439182 ops/sec; 9528.1 MB/s (4096 per op) xxh3 : 0.161 micros/op 6202872 ops/sec; 24230.0 MB/s (4096 per op) crc32c : 0.203 micros/op 4924686 ops/sec; 19237.1 MB/s (4096 per op) xxhash : 0.839 micros/op 1192388 ops/sec; 4657.8 MB/s (4096 per op) xxhash64 : 0.424 micros/op 2357391 ops/sec; 9208.6 MB/s (4096 per op) xxh3 : 0.162 micros/op 6182678 ops/sec; 24151.1 MB/s (4096 per op) As you can see, especially once warmed up, xxh3 is fastest. ### Performance macrobenchmark (PORTABLE=0 DEBUG_LEVEL=0, Broadwell processor) Test for I in `seq 1 50`; do for CHK in 0 1 2 3 4; do TEST_TMPDIR=/dev/shm/rocksdb$CHK ./db_bench -benchmarks=fillseq -memtablerep=vector -allow_concurrent_memtable_write=false -num=30000000 -checksum_type=$CHK 2>&1 | grep 'micros/op' | tee -a results-$CHK & done; wait; done Results (ops/sec) for FILE in results*; do echo -n "$FILE "; awk '{ s += $5; c++; } END { print 1.0 * s / c; }' < $FILE; done results-0 252118 # kNoChecksum results-1 251588 # kCRC32c results-2 251863 # kxxHash results-3 252016 # kxxHash64 results-4 252038 # kXXH3 Reviewed By: mrambacher Differential Revision: D31905249 Pulled By: pdillinger fbshipit-source-id: cb9b998ebe2523fc7c400eedf62124a78bf4b4d1
3 years ago
std::string garbage;
if (!SerializeEnum<ChecksumType>(checksum_type_string_map,
table_options_.checksum, &garbage)) {
return Status::InvalidArgument(
"Unrecognized ChecksumType for checksum: " +
ROCKSDB_NAMESPACE::ToString(
static_cast<uint32_t>(table_options_.checksum)));
}
return TableFactory::ValidateOptions(db_opts, cf_opts);
}
std::string BlockBasedTableFactory::GetPrintableOptions() const {
std::string ret;
ret.reserve(20000);
const int kBufferSize = 200;
char buffer[kBufferSize];
snprintf(buffer, kBufferSize, " flush_block_policy_factory: %s (%p)\n",
table_options_.flush_block_policy_factory->Name(),
static_cast<void*>(table_options_.flush_block_policy_factory.get()));
ret.append(buffer);
snprintf(buffer, kBufferSize, " cache_index_and_filter_blocks: %d\n",
table_options_.cache_index_and_filter_blocks);
ret.append(buffer);
snprintf(buffer, kBufferSize,
" cache_index_and_filter_blocks_with_high_priority: %d\n",
table_options_.cache_index_and_filter_blocks_with_high_priority);
ret.append(buffer);
snprintf(buffer, kBufferSize,
" pin_l0_filter_and_index_blocks_in_cache: %d\n",
table_options_.pin_l0_filter_and_index_blocks_in_cache);
ret.append(buffer);
snprintf(buffer, kBufferSize, " pin_top_level_index_and_filter: %d\n",
table_options_.pin_top_level_index_and_filter);
ret.append(buffer);
snprintf(buffer, kBufferSize, " index_type: %d\n",
table_options_.index_type);
ret.append(buffer);
snprintf(buffer, kBufferSize, " data_block_index_type: %d\n",
table_options_.data_block_index_type);
ret.append(buffer);
snprintf(buffer, kBufferSize, " index_shortening: %d\n",
static_cast<int>(table_options_.index_shortening));
ret.append(buffer);
snprintf(buffer, kBufferSize, " data_block_hash_table_util_ratio: %lf\n",
table_options_.data_block_hash_table_util_ratio);
ret.append(buffer);
snprintf(buffer, kBufferSize, " hash_index_allow_collision: %d\n",
table_options_.hash_index_allow_collision);
ret.append(buffer);
snprintf(buffer, kBufferSize, " checksum: %d\n", table_options_.checksum);
ret.append(buffer);
snprintf(buffer, kBufferSize, " no_block_cache: %d\n",
table_options_.no_block_cache);
ret.append(buffer);
snprintf(buffer, kBufferSize, " block_cache: %p\n",
static_cast<void*>(table_options_.block_cache.get()));
ret.append(buffer);
if (table_options_.block_cache) {
const char* block_cache_name = table_options_.block_cache->Name();
if (block_cache_name != nullptr) {
snprintf(buffer, kBufferSize, " block_cache_name: %s\n",
block_cache_name);
ret.append(buffer);
}
ret.append(" block_cache_options:\n");
ret.append(table_options_.block_cache->GetPrintableOptions());
}
snprintf(buffer, kBufferSize, " block_cache_compressed: %p\n",
static_cast<void*>(table_options_.block_cache_compressed.get()));
ret.append(buffer);
if (table_options_.block_cache_compressed) {
const char* block_cache_compressed_name =
table_options_.block_cache_compressed->Name();
if (block_cache_compressed_name != nullptr) {
snprintf(buffer, kBufferSize, " block_cache_name: %s\n",
block_cache_compressed_name);
ret.append(buffer);
}
ret.append(" block_cache_compressed_options:\n");
ret.append(table_options_.block_cache_compressed->GetPrintableOptions());
}
snprintf(buffer, kBufferSize, " persistent_cache: %p\n",
static_cast<void*>(table_options_.persistent_cache.get()));
ret.append(buffer);
if (table_options_.persistent_cache) {
snprintf(buffer, kBufferSize, " persistent_cache_options:\n");
ret.append(buffer);
ret.append(table_options_.persistent_cache->GetPrintableOptions());
}
snprintf(buffer, kBufferSize, " block_size: %" PRIu64 "\n",
table_options_.block_size);
ret.append(buffer);
snprintf(buffer, kBufferSize, " block_size_deviation: %d\n",
table_options_.block_size_deviation);
ret.append(buffer);
snprintf(buffer, kBufferSize, " block_restart_interval: %d\n",
table_options_.block_restart_interval);
ret.append(buffer);
snprintf(buffer, kBufferSize, " index_block_restart_interval: %d\n",
table_options_.index_block_restart_interval);
ret.append(buffer);
snprintf(buffer, kBufferSize, " metadata_block_size: %" PRIu64 "\n",
table_options_.metadata_block_size);
ret.append(buffer);
snprintf(buffer, kBufferSize, " partition_filters: %d\n",
table_options_.partition_filters);
ret.append(buffer);
snprintf(buffer, kBufferSize, " use_delta_encoding: %d\n",
table_options_.use_delta_encoding);
ret.append(buffer);
snprintf(buffer, kBufferSize, " filter_policy: %s\n",
table_options_.filter_policy == nullptr
? "nullptr"
: table_options_.filter_policy->Name());
ret.append(buffer);
snprintf(buffer, kBufferSize, " whole_key_filtering: %d\n",
table_options_.whole_key_filtering);
ret.append(buffer);
snprintf(buffer, kBufferSize, " verify_compression: %d\n",
table_options_.verify_compression);
ret.append(buffer);
snprintf(buffer, kBufferSize, " read_amp_bytes_per_bit: %d\n",
table_options_.read_amp_bytes_per_bit);
ret.append(buffer);
snprintf(buffer, kBufferSize, " format_version: %d\n",
table_options_.format_version);
ret.append(buffer);
snprintf(buffer, kBufferSize, " enable_index_compression: %d\n",
table_options_.enable_index_compression);
ret.append(buffer);
snprintf(buffer, kBufferSize, " block_align: %d\n",
table_options_.block_align);
ret.append(buffer);
snprintf(buffer, kBufferSize,
" max_auto_readahead_size: %" ROCKSDB_PRIszt "\n",
table_options_.max_auto_readahead_size);
ret.append(buffer);
snprintf(buffer, kBufferSize, " prepopulate_block_cache: %d\n",
static_cast<int>(table_options_.prepopulate_block_cache));
ret.append(buffer);
return ret;
}
const void* BlockBasedTableFactory::GetOptionsPtr(
const std::string& name) const {
if (name == kBlockCacheOpts()) {
if (table_options_.no_block_cache) {
return nullptr;
} else {
return table_options_.block_cache.get();
}
} else {
return TableFactory::GetOptionsPtr(name);
}
}
#ifndef ROCKSDB_LITE
// Take a default BlockBasedTableOptions "table_options" in addition to a
// map "opts_map" of option name to option value to construct the new
// BlockBasedTableOptions "new_table_options".
//
// Below are the instructions of how to config some non-primitive-typed
// options in BlockBasedTableOptions:
//
// * filter_policy:
// We currently only support the following FilterPolicy in the convenience
// functions:
// - BloomFilter: use "bloomfilter:[bits_per_key]:[use_block_based_builder]"
// to specify BloomFilter. The above string is equivalent to calling
// NewBloomFilterPolicy(bits_per_key, use_block_based_builder).
// [Example]:
// - Pass {"filter_policy", "bloomfilter:4:true"} in
// GetBlockBasedTableOptionsFromMap to use a BloomFilter with 4-bits
// per key and use_block_based_builder enabled.
//
// * block_cache / block_cache_compressed:
// We currently only support LRU cache in the GetOptions API. The LRU
// cache can be set by directly specifying its size.
// [Example]:
// - Passing {"block_cache", "1M"} in GetBlockBasedTableOptionsFromMap is
// equivalent to setting block_cache using NewLRUCache(1024 * 1024).
//
// @param table_options the default options of the output "new_table_options".
// @param opts_map an option name to value map for specifying how
// "new_table_options" should be set.
// @param new_table_options the resulting options based on "table_options"
// with the change specified in "opts_map".
// @param input_strings_escaped when set to true, each escaped characters
// prefixed by '\' in the values of the opts_map will be further converted
// back to the raw string before assigning to the associated options.
// @param ignore_unknown_options when set to true, unknown options are ignored
// instead of resulting in an unknown-option error.
// @return Status::OK() on success. Otherwise, a non-ok status indicating
// error will be returned, and "new_table_options" will be set to
// "table_options".
Status BlockBasedTableFactory::ParseOption(const ConfigOptions& config_options,
const OptionTypeInfo& opt_info,
const std::string& opt_name,
const std::string& opt_value,
void* opt_ptr) {
Status status = TableFactory::ParseOption(config_options, opt_info, opt_name,
opt_value, opt_ptr);
if (config_options.input_strings_escaped && !status.ok()) { // Got an error
// !input_strings_escaped indicates the old API, where everything is
// parsable.
if (opt_info.IsByName()) {
status = Status::OK();
}
}
return status;
}
Status GetBlockBasedTableOptionsFromString(
const BlockBasedTableOptions& table_options, const std::string& opts_str,
BlockBasedTableOptions* new_table_options) {
ConfigOptions config_options;
config_options.input_strings_escaped = false;
config_options.ignore_unknown_options = false;
config_options.invoke_prepare_options = false;
return GetBlockBasedTableOptionsFromString(config_options, table_options,
opts_str, new_table_options);
}
Status GetBlockBasedTableOptionsFromString(
const ConfigOptions& config_options,
const BlockBasedTableOptions& table_options, const std::string& opts_str,
BlockBasedTableOptions* new_table_options) {
std::unordered_map<std::string, std::string> opts_map;
Status s = StringToMap(opts_str, &opts_map);
if (!s.ok()) {
return s;
}
s = GetBlockBasedTableOptionsFromMap(config_options, table_options, opts_map,
new_table_options);
// Translate any errors (NotFound, NotSupported, to InvalidArgument
if (s.ok() || s.IsInvalidArgument()) {
return s;
} else {
return Status::InvalidArgument(s.getState());
}
}
Status GetBlockBasedTableOptionsFromMap(
const BlockBasedTableOptions& table_options,
const std::unordered_map<std::string, std::string>& opts_map,
BlockBasedTableOptions* new_table_options, bool input_strings_escaped,
bool ignore_unknown_options) {
ConfigOptions config_options;
config_options.input_strings_escaped = input_strings_escaped;
config_options.ignore_unknown_options = ignore_unknown_options;
config_options.invoke_prepare_options = false;
return GetBlockBasedTableOptionsFromMap(config_options, table_options,
opts_map, new_table_options);
}
Status GetBlockBasedTableOptionsFromMap(
const ConfigOptions& config_options,
const BlockBasedTableOptions& table_options,
const std::unordered_map<std::string, std::string>& opts_map,
BlockBasedTableOptions* new_table_options) {
assert(new_table_options);
BlockBasedTableFactory bbtf(table_options);
Status s = bbtf.ConfigureFromMap(config_options, opts_map);
if (s.ok()) {
*new_table_options = *(bbtf.GetOptions<BlockBasedTableOptions>());
} else {
*new_table_options = table_options;
}
return s;
}
#endif // !ROCKSDB_LITE
TableFactory* NewBlockBasedTableFactory(
const BlockBasedTableOptions& _table_options) {
return new BlockBasedTableFactory(_table_options);
}
const std::string BlockBasedTablePropertyNames::kIndexType =
"rocksdb.block.based.table.index.type";
const std::string BlockBasedTablePropertyNames::kWholeKeyFiltering =
"rocksdb.block.based.table.whole.key.filtering";
const std::string BlockBasedTablePropertyNames::kPrefixFiltering =
"rocksdb.block.based.table.prefix.filtering";
const std::string kHashIndexPrefixesBlock = "rocksdb.hashindex.prefixes";
const std::string kHashIndexPrefixesMetadataBlock =
"rocksdb.hashindex.metadata";
const std::string kPropTrue = "1";
const std::string kPropFalse = "0";
} // namespace ROCKSDB_NAMESPACE