andreacavalli/rocksdb
0
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Don't artificially inflate L0 score

Browse Source 
Summary:
This turns out to be pretty bad because if we prioritize L0->L1 then L1 can grow artificially large, which makes L0->L1 more and more expensive. For example:
256MB @ L0 + 256MB @ L1 --> 512MB @ L1
256MB @ L0 + 512MB @ L1 --> 768MB @ L1
256MB @ L0 + 768MB @ L1 --> 1GB @ L1

....

256MB @ L0 + 10GB @ L1 --> 10.2GB @ L1

At some point we need to start compacting L1->L2 to speed up L0->L1.

Test Plan:
The performance improvement is massive for heavy write workload. This is the benchmark I ran: https://phabricator.fb.com/P19842671. Before this change, the benchmark took 47 minutes to complete. After, the benchmark finished in 2minutes. You can see full results here: https://phabricator.fb.com/P19842674

Also, we ran this diff on MongoDB on RocksDB on one replicaset. Before the change, our initial sync was so slow that it couldn't keep up with primary writes. After the change, the import finished without any issues

Reviewers: dynamike, MarkCallaghan, rven, yhchiang, sdong

Reviewed By: sdong

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D38637
This commit is contained in:
Igor Canadi 2015-05-21 11:40:48 -07:00
parent 4cb4d546cd
commit 7a3577519f
2 changed files with 9 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
db/db_bench.cc
View File

@ -367,13 +367,19 @@ static std::vector<int> FLAGS_max_bytes_for_level_multiplier_additional_v;
DEFINE_string(max_bytes_for_level_multiplier_additional, "",
"A vector that specifies additional fanout per level");
DEFINE_int32(level0_stop_writes_trigger, 12, "Number of files in level-0"
DEFINE_int32(level0_stop_writes_trigger,
rocksdb::Options().level0_stop_writes_trigger,
"Number of files in level-0"
" that will trigger put stop.");
DEFINE_int32(level0_slowdown_writes_trigger, 8, "Number of files in level-0"
DEFINE_int32(level0_slowdown_writes_trigger,
rocksdb::Options().level0_slowdown_writes_trigger,
"Number of files in level-0"
" that will slow down writes.");
DEFINE_int32(level0_file_num_compaction_trigger, 4, "Number of files in level-0"
DEFINE_int32(level0_file_num_compaction_trigger,
rocksdb::Options().level0_file_num_compaction_trigger,
"Number of files in level-0"
" when compactions start");
static bool ValidateInt32Percent(const char* flagname, int32_t value) {

7
db/version_set.cc
View File

@ -1019,13 +1019,6 @@ void VersionStorageInfo::ComputeCompactionScore(
if (compaction_style_ == kCompactionStyleFIFO) {
score = static_cast<double>(total_size) /
compaction_options_fifo.max_table_files_size;
} else if (num_sorted_runs >=
mutable_cf_options.level0_stop_writes_trigger) {
// If we are slowing down writes, then we better compact that first
score = 1000000;
} else if (num_sorted_runs >=
mutable_cf_options.level0_slowdown_writes_trigger) {
score = 10000;
} else {
score = static_cast<double>(num_sorted_runs) /
mutable_cf_options.level0_file_num_compaction_trigger;