rocksdb/table/block_based
Peter Dillinger fc9d4071f0 Fast path for detecting unchanged prefix_extractor (#9407)
Summary:
Fixes a major performance regression in 6.26, where
extra CPU is spent in SliceTransform::AsString when reads involve
a prefix_extractor (Get, MultiGet, Seek). Common case performance
is now better than 6.25.

This change creates a "fast path" for verifying that the current prefix
extractor is unchanged and compatible with what was used to
generate a table file. This fast path detects the common case by
pointer comparison on the current prefix_extractor and a "known
good" prefix extractor (if applicable) that is saved at the time the
table reader is opened. The "known good" prefix extractor is saved
as another shared_ptr copy (in an existing field, however) to ensure
the pointer is not recycled.

When the prefix_extractor has changed to a different instance but
same compatible configuration (rare, odd), performance is still a
regression compared to 6.25, but this is likely acceptable because
of the oddity of such a case. The performance of incompatible
prefix_extractor is essentially unchanged.

Also fixed a minor case (ForwardIterator) where a prefix_extractor
could be used via a raw pointer after being freed as a shared_ptr,
if replaced via SetOptions.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/9407

Test Plan:
## Performance
Populate DB with `TEST_TMPDIR=/dev/shm/rocksdb ./db_bench -benchmarks=fillrandom -num=10000000 -disable_wal=1 -write_buffer_size=10000000 -bloom_bits=16 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=0 -prefix_size=12`

Running head-to-head comparisons simultaneously with `TEST_TMPDIR=/dev/shm/rocksdb ./db_bench -use_existing_db -readonly -benchmarks=seekrandom -num=10000000 -duration=20 -disable_wal=1 -bloom_bits=16 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=0 -prefix_size=12`

Below each is compared by ops/sec vs. baseline which is version 6.25 (multiple baseline runs because of variable machine load)

v6.26: 4833 vs. 6698 (<- major regression!)
v6.27: 4737 vs. 6397 (still)
New: 6704 vs. 6461 (better than baseline in common case)
Disabled fastpath: 4843 vs. 6389 (e.g. if prefix extractor instance changes but is still compatible)
Changed prefix size (no usable filter) in new: 787 vs. 5927
Changed prefix size (no usable filter) in new & baseline: 773 vs. 784

Reviewed By: mrambacher

Differential Revision: D33677812

Pulled By: pdillinger

fbshipit-source-id: 571d9711c461fb97f957378a061b7e7dbc4d6a76
2022-01-21 11:37:46 -08:00
..
binary_search_index_reader.cc Separate internal and user key comparators in BlockIter (#6944) 2020-07-07 17:26:16 -07:00
binary_search_index_reader.h Extend Get/MultiGet deadline support to table open (#6982) 2020-06-29 14:53:17 -07:00
block_based_filter_block_test.cc Add table properties for number of entries added to filters (#8323) 2021-05-21 17:11:32 -07:00
block_based_filter_block.cc Deallocate payload of BlockBasedTableBuilder::Rep::FilterBlockBuilder earlier for Full/PartitionedFilter (#9070) 2021-11-04 13:35:38 -07:00
block_based_filter_block.h Deallocate payload of BlockBasedTableBuilder::Rep::FilterBlockBuilder earlier for Full/PartitionedFilter (#9070) 2021-11-04 13:35:38 -07:00
block_based_table_builder.cc Enable db_test running in Centos 32 bit OS and Alpine 32 bit OS (#9294) 2022-01-14 11:58:18 -08:00
block_based_table_builder.h Fix segmentation fault in table_options.prepopulate_block_cache when used with partition_filters (#9263) 2021-12-08 12:44:38 -08:00
block_based_table_factory.cc Enable db_test running in Centos 32 bit OS and Alpine 32 bit OS (#9294) 2022-01-14 11:58:18 -08:00
block_based_table_factory.h Cleanup includes in dbformat.h (#8930) 2021-09-29 04:04:40 -07:00
block_based_table_iterator.cc Reuse internal auto readhead_size at each Level (expect L0) for Iterations (#9056) 2021-11-10 16:20:04 -08:00
block_based_table_iterator.h Fix bug in rocksdb internal automatic prefetching (#9234) 2021-11-30 22:53:10 -08:00
block_based_table_reader_impl.h New stable, fixed-length cache keys (#9126) 2021-12-16 17:15:13 -08:00
block_based_table_reader_test.cc Cleanup includes in dbformat.h (#8930) 2021-09-29 04:04:40 -07:00
block_based_table_reader.cc Fast path for detecting unchanged prefix_extractor (#9407) 2022-01-21 11:37:46 -08:00
block_based_table_reader.h Fast path for detecting unchanged prefix_extractor (#9407) 2022-01-21 11:37:46 -08:00
block_builder.cc Improve data block construction performance (#9040) 2021-10-19 12:36:21 -07:00
block_builder.h Improve data block construction performance (#9040) 2021-10-19 12:36:21 -07:00
block_like_traits.h fix lru caching test and fix reference binding to null pointer (#8326) 2021-05-24 08:37:00 -07:00
block_prefetcher.cc Improve / clean up meta block code & integrity (#9163) 2021-11-18 11:43:44 -08:00
block_prefetcher.h Fix bug in rocksdb internal automatic prefetching (#9234) 2021-11-30 22:53:10 -08:00
block_prefix_index.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
block_prefix_index.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
block_test.cc Improve / clean up meta block code & integrity (#9163) 2021-11-18 11:43:44 -08:00
block_type.h Fix and detect headers with missing dependencies (#8893) 2021-09-10 10:00:26 -07:00
block.cc Add NewMetaDataIterator method (#8692) 2021-12-21 11:32:49 -08:00
block.h Add NewMetaDataIterator method (#8692) 2021-12-21 11:32:49 -08:00
cachable_entry.h Parallelize secondary cache lookup in MultiGet (#8405) 2021-06-18 09:35:59 -07:00
data_block_footer.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
data_block_footer.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
data_block_hash_index_test.cc Fast path for detecting unchanged prefix_extractor (#9407) 2022-01-21 11:37:46 -08:00
data_block_hash_index.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
data_block_hash_index.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
filter_block_reader_common.cc Parallelize secondary cache lookup in MultiGet (#8405) 2021-06-18 09:35:59 -07:00
filter_block_reader_common.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
filter_block.h Account Bloom/Ribbon filter construction memory in global memory limit (#9073) 2021-11-18 09:42:20 -08:00
filter_policy_internal.h Add Bloom/Ribbon hybrid API support (#8679) 2021-08-20 18:00:16 -07:00
filter_policy.cc Make some FilterPolicy deprecations more clear (#9403) 2022-01-19 18:12:10 -08:00
flush_block_policy.cc Restore Regex support for ObjectLibrary::Register, rename new APIs to allow old one to be deprecated in the future (#9362) 2022-01-11 06:33:48 -08:00
flush_block_policy.h Make FlushBlockPolicyFactory into a Customizable class (#8432) 2021-07-12 09:04:59 -07:00
full_filter_block_test.cc Add table properties for number of entries added to filters (#8323) 2021-05-21 17:11:32 -07:00
full_filter_block.cc Deallocate payload of BlockBasedTableBuilder::Rep::FilterBlockBuilder earlier for Full/PartitionedFilter (#9070) 2021-11-04 13:35:38 -07:00
full_filter_block.h Account Bloom/Ribbon filter construction memory in global memory limit (#9073) 2021-11-18 09:42:20 -08:00
hash_index_reader.cc Make ImmutableOptions struct that inherits from ImmutableCFOptions and ImmutableDBOptions (#8262) 2021-05-05 14:00:17 -07:00
hash_index_reader.h Extend Get/MultiGet deadline support to table open (#6982) 2020-06-29 14:53:17 -07:00
index_builder.cc Add (& fix) some simple source code checks (#8821) 2021-09-07 21:19:27 -07:00
index_builder.h Make db_basic_test pass assert status checked (#7452) 2020-09-29 09:49:04 -07:00
index_reader_common.cc Parallelize secondary cache lookup in MultiGet (#8405) 2021-06-18 09:35:59 -07:00
index_reader_common.h Divide block_based_table_reader.cc (#6527) 2020-03-12 21:41:50 -07:00
mock_block_based_table.h Make ImmutableOptions struct that inherits from ImmutableCFOptions and ImmutableDBOptions (#8262) 2021-05-05 14:00:17 -07:00
parsed_full_filter_block.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
parsed_full_filter_block.h Use new Insert and Lookup APIs in table reader to support secondary cache (#8315) 2021-05-21 18:29:12 -07:00
partitioned_filter_block_test.cc More refactoring ahead of footer & meta changes (#9240) 2021-12-10 08:13:26 -08:00
partitioned_filter_block.cc Improve / clean up meta block code & integrity (#9163) 2021-11-18 11:43:44 -08:00
partitioned_filter_block.h Deallocate payload of BlockBasedTableBuilder::Rep::FilterBlockBuilder earlier for Full/PartitionedFilter (#9070) 2021-11-04 13:35:38 -07:00
partitioned_index_iterator.cc Reuse internal auto readhead_size at each Level (expect L0) for Iterations (#9056) 2021-11-10 16:20:04 -08:00
partitioned_index_iterator.h Fix bug in rocksdb internal automatic prefetching (#9234) 2021-11-30 22:53:10 -08:00
partitioned_index_reader.cc Improve / clean up meta block code & integrity (#9163) 2021-11-18 11:43:44 -08:00
partitioned_index_reader.h Clarify caching behavior for index and filter partitions (#9068) 2021-10-27 17:23:04 -07:00
reader_common.cc Improve / clean up meta block code & integrity (#9163) 2021-11-18 11:43:44 -08:00
reader_common.h Bring the Configurable options together (#5753) 2020-09-14 17:01:01 -07:00
uncompression_dict_reader.cc Cleanup includes in dbformat.h (#8930) 2021-09-29 04:04:40 -07:00
uncompression_dict_reader.h Extend Get/MultiGet deadline support to table open (#6982) 2020-06-29 14:53:17 -07:00