rocksdb/db
Andrew Kryczka 177f8bd063 Bound L0->Lbase fanout in dynamic leveled compaction (#7325)
Summary:
L0 score is based on size target and number of files. The size target
used is `max_bytes_for_level_base`. However, the base level's size can
dynamically expand in write burst mode. In fact, it can expand so much
that L0->Lbase becomes the highest fanout in target sizes. This doesn't
make sense from an efficiency perspective, so this PR bounds the
L0->Lbase fanout to the smoothed level multiplier. The L0 scoring based
on file count remains unchanged.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/7325

Test Plan:
contrived benchmark that exhibits the problem:

```
$ TEST_TMPDIR=/data/users/andrewkr/ ./db_bench -benchmarks=filluniquerandom,readrandom -write_buffer_size=1048576 -target_file_size_base=1048576 -max_bytes_for_level_base=4194304 -level0_file_num_compaction_trigger=4 -level_compaction_dynamic_level_bytes=true -compression_type=none -max_background_jobs=12 -rate_limiter_bytes_per_sec=104857600 -benchmark_write_rate_limit=10485760 -num=100000000
```

Results:

- "Burst W-Amp" is the write-amp near the end of the fillrandom benchmark
- "Total W-Amp" is the write-amp after readrandom has run a while and all levels no longer need compaction

Branch | Burst W-Amp | Total W-Amp | fillrandom (MB/s)
-- | -- | -- | --
master | 20.2 | 21.5 | 4.7
dynamic-l0-score | 12.6 | 14.1 | 7.2

Reviewed By: siying

Differential Revision: D23412935

Pulled By: ajkr

fbshipit-source-id: f91f2067188e432dd39deab02f1c56f195057a0e
2020-09-01 19:34:01 -07:00
..
blob Log info about generated blob files in BlobFileBuilder (#7324) 2020-08-31 13:24:12 -07:00
compaction Pass SST file checksum information through OnTableFileCreated (#7108) 2020-08-25 10:46:11 -07:00
db_impl Add some simulator cache and block tracer tests to ASSERT_STATUS_CHECKED (#7305) 2020-08-24 16:43:31 -07:00
arena_wrapped_db_iter.cc dedup ReadOptions in iterator hierarchy (#7210) 2020-08-03 15:23:04 -07:00
arena_wrapped_db_iter.h dedup ReadOptions in iterator hierarchy (#7210) 2020-08-03 15:23:04 -07:00
builder.cc Pass SST file checksum information through OnTableFileCreated (#7108) 2020-08-25 10:46:11 -07:00
builder.h Store DB identity and DB session ID in SST files (#6983) 2020-06-17 10:57:40 -07:00
c_test.c Add getters for the read options to the C API (#7289) 2020-08-20 16:36:19 -07:00
c.cc Add getters for the read options to the C API (#7289) 2020-08-20 16:36:19 -07:00
column_family_test.cc column_family_test: fix a data race related to sleeping task (#7150) 2020-07-20 14:19:48 -07:00
column_family.cc Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
column_family.h Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
compact_files_test.cc Replace reinterpret_cast with static_cast_with_check (#7067) 2020-07-02 19:25:41 -07:00
compacted_db_impl.cc Introduce a global StatsDumpScheduler for stats dumping (#7223) 2020-08-14 20:12:44 -07:00
compacted_db_impl.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
comparator_db_test.cc More Makefile Cleanup (#7097) 2020-07-09 14:35:17 -07:00
convenience.cc sst_dump to reduce number of file reads (#6836) 2020-05-12 18:23:33 -07:00
corruption_test.cc Upgrade tool chain (#7251) 2020-08-12 19:30:00 -07:00
cuckoo_table_db_test.cc Replace reinterpret_cast with static_cast_with_check (#7067) 2020-07-02 19:25:41 -07:00
db_basic_test.cc Parameterize DBBasicTest.CompactBetweenSnapshots (#7301) 2020-08-25 15:42:11 -07:00
db_block_cache_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_bloom_filter_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_compaction_filter_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_compaction_test.cc Eliminates a no-op compaction upon snapshot release when disabling auto compactions (#7267) 2020-08-24 22:06:45 -07:00
db_dynamic_level_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_encryption_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_filesnapshot.cc First step towards handling MANIFEST write error (#6949) 2020-06-24 19:07:08 -07:00
db_flush_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_info_dumper.cc Add some simulator cache and block tracer tests to ASSERT_STATUS_CHECKED (#7305) 2020-08-24 16:43:31 -07:00
db_info_dumper.h Add a DB Session ID (#6959) 2020-06-15 10:47:02 -07:00
db_inplace_update_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_io_failure_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_iter_stress_test.cc Test CircleCI with CLANG-10 (#7025) 2020-06-24 16:22:49 -07:00
db_iter_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
db_iter.cc Add some simulator cache and block tracer tests to ASSERT_STATUS_CHECKED (#7305) 2020-08-24 16:43:31 -07:00
db_iter.h ReadOptions.iter_start_ts should support tombstones (#7178) 2020-08-04 18:52:08 -07:00
db_iterator_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_log_iter_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_logical_block_size_cache_test.cc Get block size only in direct IO mode (#6522) 2020-03-20 15:26:10 -07:00
db_memtable_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_merge_operand_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_merge_operator_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_options_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_properties_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_range_del_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_sst_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_statistics_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_table_properties_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_tailing_iter_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_test2.cc Get() to fail with underlying failures in PartitionIndexReader::CacheDependencies() (#7297) 2020-08-25 19:01:05 -07:00
db_test_util.cc Add some simulator cache and block tracer tests to ASSERT_STATUS_CHECKED (#7305) 2020-08-24 16:43:31 -07:00
db_test_util.h Get() to fail with underlying failures in PartitionIndexReader::CacheDependencies() (#7297) 2020-08-25 19:01:05 -07:00
db_test.cc Add unit test for max_write_buffer_size_to_maintain (#7311) 2020-08-28 17:38:05 -07:00
db_universal_compaction_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_wal_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_with_timestamp_basic_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_with_timestamp_compaction_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
db_write_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
dbformat_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
dbformat.cc Add a new EntryType for deletion with timestamp (#7195) 2020-08-17 16:26:06 -07:00
dbformat.h Separate internal and user key comparators in BlockIter (#6944) 2020-07-07 17:26:16 -07:00
deletefile_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
error_handler_fs_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
error_handler.cc Auto resume the DB from Retryable IO Error (#6765) 2020-07-15 11:03:58 -07:00
error_handler.h Add EnvTestWithParam::OptionsTest to the ASSERT_STATUS_CHECKED passes (#7283) 2020-08-20 19:18:35 -07:00
event_helpers.cc Pass SST file checksum information through OnTableFileCreated (#7108) 2020-08-25 10:46:11 -07:00
event_helpers.h Pass SST file checksum information through OnTableFileCreated (#7108) 2020-08-25 10:46:11 -07:00
experimental.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
external_sst_file_basic_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
external_sst_file_ingestion_job.cc Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
external_sst_file_ingestion_job.h Store FSSequentialFilePtr object in SequenceFileReader (#7190) 2020-08-18 16:20:54 -07:00
external_sst_file_test.cc Disable fsync in some ExternalSSTFileTest tests (#7303) 2020-08-24 11:26:09 -07:00
fault_injection_test.cc More Makefile Cleanup (#7097) 2020-07-09 14:35:17 -07:00
file_indexer_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
file_indexer.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
file_indexer.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
filename_test.cc Add more tests to ASSERT_STATUS_CHECKED (#7211) 2020-08-06 17:19:41 -07:00
flush_job_test.cc Build blob file reader/writer classes in LITE mode as well (#7272) 2020-08-17 15:19:05 -07:00
flush_job.cc Fix+clean up handling of mock sleeps (#7101) 2020-08-11 12:41:30 -07:00
flush_job.h Store DB identity and DB session ID in SST files (#6983) 2020-06-17 10:57:40 -07:00
flush_scheduler.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
flush_scheduler.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
forward_iterator_bench.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
forward_iterator.cc make L0 index/filter pinned memory usage predictable (#6911) 2020-06-09 16:51:23 -07:00
forward_iterator.h Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621) 2020-04-15 17:40:44 -07:00
import_column_family_job.cc Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
import_column_family_job.h Store FSSequentialFilePtr object in SequenceFileReader (#7190) 2020-08-18 16:20:54 -07:00
import_column_family_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
internal_stats.cc First step towards handling MANIFEST write error (#6949) 2020-06-24 19:07:08 -07:00
internal_stats.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
job_context.h Expose the set of live blob files from Version/VersionSet (#6785) 2020-05-04 15:08:13 -07:00
listener_test.cc Pass SST file checksum information through OnTableFileCreated (#7108) 2020-08-25 10:46:11 -07:00
log_format.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
log_reader.cc Real fix for race in backup custom checksum checking (#7309) 2020-08-26 10:39:20 -07:00
log_reader.h Real fix for race in backup custom checksum checking (#7309) 2020-08-26 10:39:20 -07:00
log_test.cc Revert "Update googletest from 1.8.1 to 1.10.0 (#6808)" (#6923) 2020-06-03 15:55:03 -07:00
log_writer.cc Fail recovery when MANIFEST record checksum mismatch (#6996) 2020-06-18 10:09:12 -07:00
log_writer.h Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487) 2020-03-27 16:04:43 -07:00
logs_with_prep_tracker.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
logs_with_prep_tracker.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
lookup_key.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
malloc_stats.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
malloc_stats.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
manual_compaction_test.cc Skip high levels with no key falling in the range in CompactRange (#6482) 2020-03-04 20:15:25 -08:00
memtable_list_test.cc Store FileSystemPtr object that contains FileSystem ptr (#7180) 2020-08-12 17:31:23 -07:00
memtable_list.cc Bug Fix for memtables not trimmed down. (#7296) 2020-08-21 13:29:05 -07:00
memtable_list.h Bug Fix for memtables not trimmed down. (#7296) 2020-08-21 13:29:05 -07:00
memtable.cc Clean up InternalIterator upper bound logic a little bit (#7200) 2020-08-05 10:44:57 -07:00
memtable.h return timestamp from get (#6409) 2020-03-02 16:01:00 -08:00
merge_context.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
merge_helper_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
merge_helper.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
merge_helper.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
merge_operator.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
merge_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
obsolete_files_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
options_file_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
perf_context_test.cc C++20 compatibility (#6697) 2020-04-20 13:24:25 -07:00
pinned_iterators_manager.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
plain_table_db_test.cc More Makefile Cleanup (#7097) 2020-07-09 14:35:17 -07:00
pre_release_callback.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
prefix_test.cc Fix cmake build on MacOS (#7205) 2020-08-10 10:48:05 -07:00
range_del_aggregator_bench.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
range_del_aggregator_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
range_del_aggregator.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
range_del_aggregator.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
range_tombstone_fragmenter_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
range_tombstone_fragmenter.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
range_tombstone_fragmenter.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
read_callback.h Get() with timestamp should respect snapshot (#7227) 2020-08-14 19:20:58 -07:00
repair_test.cc Whole DBTest to skip fsync (#7274) 2020-08-17 18:42:25 -07:00
repair.cc Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
snapshot_checker.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
snapshot_impl.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
snapshot_impl.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
table_cache.cc Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
table_cache.h Store FSRandomAccessPtr object in RandomAccessFileReader (#7192) 2020-08-27 11:21:52 -07:00
table_properties_collector_test.cc Revert "Update googletest from 1.8.1 to 1.10.0 (#6808)" (#6923) 2020-06-03 15:55:03 -07:00
table_properties_collector.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
table_properties_collector.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
transaction_log_impl.cc Store FSSequentialFilePtr object in SequenceFileReader (#7190) 2020-08-18 16:20:54 -07:00
transaction_log_impl.h Store FileSystemPtr object that contains FileSystem ptr (#7180) 2020-08-12 17:31:23 -07:00
trim_history_scheduler.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
trim_history_scheduler.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
version_builder_test.cc Clean up blob files based on the linked SST set (#7001) 2020-06-30 15:31:21 -07:00
version_builder.cc Remove redundant consistency check from VersionStorageInfo::AddFile (#7237) 2020-08-11 09:23:17 -07:00
version_builder.h make L0 index/filter pinned memory usage predictable (#6911) 2020-06-09 16:51:23 -07:00
version_edit_handler.cc Fail recovery when MANIFEST record checksum mismatch (#6996) 2020-06-18 10:09:12 -07:00
version_edit_handler.h Fail recovery when MANIFEST record checksum mismatch (#6996) 2020-06-18 10:09:12 -07:00
version_edit_test.cc Define WAL related classes to be used in VersionEdit and VersionSet (#7164) 2020-08-05 16:34:38 -07:00
version_edit.cc Build blob file reader/writer classes in LITE mode as well (#7272) 2020-08-17 15:19:05 -07:00
version_edit.h BackupEngine supports custom file checksums (#7085) 2020-08-12 13:31:09 -07:00
version_set_test.cc Store FileSystemPtr object that contains FileSystem ptr (#7180) 2020-08-12 17:31:23 -07:00
version_set.cc Bound L0->Lbase fanout in dynamic leveled compaction (#7325) 2020-09-01 19:34:01 -07:00
version_set.h Store FileSystemPtr object that contains FileSystem ptr (#7180) 2020-08-12 17:31:23 -07:00
wal_edit_test.cc Track WAL in MANIFEST: minor udpates (#7282) 2020-08-20 15:12:00 -07:00
wal_edit.cc Track WAL in MANIFEST: minor udpates (#7282) 2020-08-20 15:12:00 -07:00
wal_edit.h Define WAL related classes to be used in VersionEdit and VersionSet (#7164) 2020-08-05 16:34:38 -07:00
wal_manager_test.cc Store FileSystemPtr object that contains FileSystem ptr (#7180) 2020-08-12 17:31:23 -07:00
wal_manager.cc Store FSSequentialFilePtr object in SequenceFileReader (#7190) 2020-08-18 16:20:54 -07:00
wal_manager.h Store FileSystemPtr object that contains FileSystem ptr (#7180) 2020-08-12 17:31:23 -07:00
write_batch_base.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_batch_internal.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_batch_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_batch.cc Add some simulator cache and block tracer tests to ASSERT_STATUS_CHECKED (#7305) 2020-08-24 16:43:31 -07:00
write_callback_test.cc Divide WriteCallbackTest.WriteWithCallbackTest (#7037) 2020-06-30 12:31:30 -07:00
write_callback.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_controller_test.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_controller.cc Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_controller.h Replace namespace name "rocksdb" with ROCKSDB_NAMESPACE (#6433) 2020-02-20 12:09:57 -08:00
write_thread.cc fix some spelling typos (#6464) 2020-02-28 14:14:03 -08:00
write_thread.h Add EnvTestWithParam::OptionsTest to the ASSERT_STATUS_CHECKED passes (#7283) 2020-08-20 19:18:35 -07:00