andreacavalli/rocksdb
0
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
499ebb3ab5
rocksdb/db
History
Maysam Yabandeh 499ebb3ab5 Optimize for serial commits in 2PC 
Summary:
Throughput: 46k tps in our sysbench settings (filling the details later)

The idea is to have the simplest change that gives us a reasonable boost
in 2PC throughput.

Major design changes:
1. The WAL file internal buffer is not flushed after each write. Instead
it is flushed before critical operations (WAL copy via fs) or when
FlushWAL is called by MySQL. Flushing the WAL buffer is also protected
via mutex_.
2. Use two sequence numbers: last seq, and last seq for write. Last seq
is the last visible sequence number for reads. Last seq for write is the
next sequence number that should be used to write to WAL/memtable. This
allows to have a memtable write be in parallel to WAL writes.
3. BatchGroup is not used for writes. This means that we can have
parallel writers which changes a major assumption in the code base. To
accommodate for that i) allow only 1 WriteImpl that intends to write to
memtable via mem_mutex_--which is fine since in 2PC almost all of the memtable writes
come via group commit phase which is serial anyway, ii) make all the
parts in the code base that assumed to be the only writer (via
EnterUnbatched) to also acquire mem_mutex_, iii) stat updates are
protected via a stat_mutex_.

Note: the first commit has the approach figured out but is not clean.
Submitting the PR anyway to get the early feedback on the approach. If
we are ok with the approach I will go ahead with this updates:
0) Rebase with Yi's pipelining changes
1) Currently batching is disabled by default to make sure that it will be
consistent with all unit tests. Will make this optional via a config.
2) A couple of unit tests are disabled. They need to be updated with the
serial commit of 2PC taken into account.
3) Replacing BatchGroup with mem_mutex_ got a bit ugly as it requires
releasing mutex_ beforehand (the same way EnterUnbatched does). This
needs to be cleaned up.
Closes https://github.com/facebook/rocksdb/pull/2345

Differential Revision: D5210732

Pulled By: maysamyabandeh

fbshipit-source-id: 78653bd95a35cd1e831e555e0e57bdfd695355a4
2017-06-24 14:11:29 -07:00
..
builder.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
builder.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
c_test.c support merge and delete in file ingestion 2017-05-26 12:11:21 -07:00
c.cc support merge and delete in file ingestion 2017-05-26 12:11:21 -07:00
column_family_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
column_family.cc Fix Data Race Between CreateColumnFamily() and GetAggregatedIntProperty() 2017-06-22 15:56:47 -07:00
column_family.h Fix Data Race Between CreateColumnFamily() and GetAggregatedIntProperty() 2017-06-22 15:56:47 -07:00
compact_files_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compacted_db_impl.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compacted_db_impl.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compaction_iteration_stats.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compaction_iterator_test.cc Support ingest_behind for IngestExternalFile 2017-05-17 11:42:42 -07:00
compaction_iterator.cc Fix RocksDB Lite build with CLANG 2017-06-12 06:41:27 -07:00
compaction_iterator.h Fix RocksDB Lite build with CLANG 2017-06-12 06:41:27 -07:00
compaction_job_stats_test.cc New API for background work in single thread pool 2017-05-23 11:12:27 -07:00
compaction_job_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
compaction_job.cc Introduce OnBackgroundError callback 2017-06-22 19:41:50 -07:00
compaction_job.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compaction_picker_test.cc account for L0 size in estimated compaction bytes 2017-06-01 17:56:59 -07:00
compaction_picker_universal.cc fixed typo 2017-06-05 11:27:34 -07:00
compaction_picker_universal.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compaction_picker.cc Support ingest_behind for IngestExternalFile 2017-05-17 11:42:42 -07:00
compaction_picker.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
compaction.cc Avoid calling fallocate with UINT64_MAX 2017-05-04 17:43:22 -07:00
compaction.h Sample number of reads per SST file 2017-06-12 07:12:08 -07:00
comparator_db_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
convenience.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
corruption_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
cuckoo_table_db_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_basic_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
db_block_cache_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_bloom_filter_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
db_compaction_filter_test.cc Fix interaction between CompactionFilter::Decision::kRemoveAndSkipUnt… 2017-06-02 15:11:38 -07:00
db_compaction_test.cc fixed typo 2017-06-13 16:58:01 -07:00
db_dynamic_level_test.cc Improve write buffer manager (and allow the size to be tracked in block cache) 2017-06-02 14:26:56 -07:00
db_filesnapshot.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_flush_test.cc New API for background work in single thread pool 2017-05-23 11:12:27 -07:00
db_impl_compaction_flush.cc Introduce OnBackgroundError callback 2017-06-22 19:41:50 -07:00
db_impl_debug.cc Introduce max_background_jobs mutable option 2017-05-24 11:29:08 -07:00
db_impl_experimental.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_impl_files.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_impl_open.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_impl_readonly.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_impl_readonly.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_impl_write.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_impl.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_impl.h Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_info_dumper.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_info_dumper.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_inplace_update_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_io_failure_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_iter_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
db_iter.cc Histogram of number of merge operands 2017-05-31 07:41:44 -07:00
db_iter.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_iterator_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
db_log_iter_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_memtable_test.cc Improve write buffer manager (and allow the size to be tracked in block cache) 2017-06-02 14:26:56 -07:00
db_merge_operator_test.cc support merge and delete in file ingestion 2017-05-26 12:11:21 -07:00
db_options_test.cc Introduce max_background_jobs mutable option 2017-05-24 11:29:08 -07:00
db_properties_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
db_range_del_test.cc Disable DBRangeDelTest::TailingIteratorRangeTombstoneUnsupported for ubsan 2017-06-08 12:43:01 -07:00
db_sst_test.cc Limit trash directory to be 25% of total DB 2017-06-12 16:57:21 -07:00
db_statistics_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_table_properties_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_tailing_iter_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
db_test2.cc Call RateLimiter for compaction reads 2017-06-13 14:56:46 -07:00
db_test_util.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_test_util.h Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_test.cc WriteOptions.low_pri which can throttle low pri writes if needed 2017-06-05 15:02:35 -07:00
db_universal_compaction_test.cc Fixed some spelling mistakes 2017-05-17 23:12:36 -07:00
db_wal_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
db_write_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
dbformat_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
dbformat.cc do not read next datablock if upperbound is reached 2017-05-05 23:20:01 -07:00
dbformat.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
deletefile_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
event_helpers.cc Introduce OnBackgroundError callback 2017-06-22 19:41:50 -07:00
event_helpers.h Introduce OnBackgroundError callback 2017-06-22 19:41:50 -07:00
experimental.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
external_sst_file_basic_test.cc Support ingest file when range deletions exist 2017-05-31 13:57:19 -07:00
external_sst_file_ingestion_job.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
external_sst_file_ingestion_job.h fixed typo 2017-06-05 11:27:34 -07:00
external_sst_file_test.cc support merge and delete in file ingestion 2017-05-26 12:11:21 -07:00
fault_injection_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
file_indexer_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
file_indexer.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
file_indexer.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
filename_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
flush_job_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
flush_job.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
flush_job.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
flush_scheduler.cc New WriteImpl to pipeline WAL/memtable write 2017-05-19 14:26:42 -07:00
flush_scheduler.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
forward_iterator_bench.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
forward_iterator.cc do not read next datablock if upperbound is reached 2017-05-05 23:20:01 -07:00
forward_iterator.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
internal_stats.cc Sample number of reads per SST file 2017-06-12 07:12:08 -07:00
internal_stats.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
job_context.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
listener_test.cc Introduce OnBackgroundError callback 2017-06-22 19:41:50 -07:00
log_format.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
log_reader.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
log_reader.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
log_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
log_writer.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
log_writer.h Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
malloc_stats.cc db: avoid #includeing malloc and jemalloc simultaneously 2017-05-31 22:43:02 -07:00
malloc_stats.h db: avoid #includeing malloc and jemalloc simultaneously 2017-05-31 22:43:02 -07:00
managed_iterator.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
managed_iterator.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
manual_compaction_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
memtable_list_test.cc Pass CF ID to MemTableRepFactory 2017-06-02 12:12:06 -07:00
memtable_list.cc Fix TSAN: avoid arena mode with range deletions 2017-06-01 22:26:49 -07:00
memtable_list.h Fix TSAN: avoid arena mode with range deletions 2017-06-01 22:26:49 -07:00
memtable.cc Improve write buffer manager (and allow the size to be tracked in block cache) 2017-06-02 14:26:56 -07:00
memtable.h Improve write buffer manager (and allow the size to be tracked in block cache) 2017-06-02 14:26:56 -07:00
merge_context.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
merge_helper_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
merge_helper.cc Histogram of number of merge operands 2017-05-31 07:41:44 -07:00
merge_helper.h Histogram of number of merge operands 2017-05-31 07:41:44 -07:00
merge_operator.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
merge_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
options_file_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
perf_context_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
pinned_iterators_manager.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
plain_table_db_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
prefix_test.cc using ThreadLocalPtr to hide ROCKSDB_SUPPORT_THREAD_LOCAL from public… 2017-06-02 17:26:19 -07:00
range_del_aggregator_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
range_del_aggregator.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
range_del_aggregator.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
repair_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
repair.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
snapshot_impl.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
snapshot_impl.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
table_cache.cc fix coredump for release nullptr 2017-06-19 11:41:38 -07:00
table_cache.h Call RateLimiter for compaction reads 2017-06-13 14:56:46 -07:00
table_properties_collector_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
table_properties_collector.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
table_properties_collector.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
transaction_log_impl.cc disable direct reads for log and manifest and add direct io to tests 2017-05-22 18:41:28 -07:00
transaction_log_impl.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
version_builder_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
version_builder.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
version_builder.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
version_edit_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
version_edit.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
version_edit.h Sample number of reads per SST file 2017-06-12 07:12:08 -07:00
version_set_test.cc Set lower-bound on dynamic level sizes 2017-05-04 18:16:12 -07:00
version_set.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
version_set.h Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
wal_manager_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
wal_manager.cc disable direct reads for log and manifest and add direct io to tests 2017-05-22 18:41:28 -07:00
wal_manager.h Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
write_batch_base.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
write_batch_internal.h Fixing blob db sequence number handling 2017-05-31 10:56:45 -07:00
write_batch_test.cc Pass CF ID to MemTableRepFactory 2017-06-02 12:12:06 -07:00
write_batch.cc Fixing blob db sequence number handling 2017-05-31 10:56:45 -07:00
write_callback_test.cc Optimize for serial commits in 2PC 2017-06-24 14:11:29 -07:00
write_callback.h Updated all copyright headers to the new format. 2016-02-09 15:12:00 -08:00
write_controller_test.cc Add GPLv2 as an alternative license. 2017-04-27 18:06:12 -07:00
write_controller.cc WriteOptions.low_pri which can throttle low pri writes if needed 2017-06-05 15:02:35 -07:00
write_controller.h WriteOptions.low_pri which can throttle low pri writes if needed 2017-06-05 15:02:35 -07:00
write_thread.cc New WriteImpl to pipeline WAL/memtable write 2017-05-19 14:26:42 -07:00
write_thread.h New WriteImpl to pipeline WAL/memtable write 2017-05-19 14:26:42 -07:00