Avoid cache lookups for range deletion meta-block #1801
Conversation
|
@ajkr has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
|
What's the reason to do Ref() only instead of normal lookup? |
|
Oh, Lookup() was consuming a lot of CPU (~5% in my profiling) for computing hash to lookup in handle table. But it's not necessary when we already have a handle. |
|
maybe i've been desensitized to overengineering, but its complexity is pretty in line with my expectations. |
ajkr
force-pushed
the
del-range-reuse-cache-handle
branch
from
63755c2 to
56b2629
Compare
|
@ajkr updated the pull request - view changes - changes since last import |
|
@ajkr updated the pull request - view changes - changes since last import |
|
@ajkr updated the pull request - view changes - changes since last import |
|
@ajkr has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
I added the Cache::Ref() function a couple weeks ago (#1761) to make this feature possible. Like other meta-blocks, rep_->range_del_entry holds a cache handle to pin the range deletion block in uncompressed block cache for the duration of the table reader's lifetime. We can reuse this cache handle to create an iterator over this meta-block without any cache lookup. Ref() is used to increment the cache handle's refcount in case the returned iterator outlives the table reader.
Test Plan:
Measurements:
Commands:
./db_bench -benchmarks=fillrandom -write_buffer_size=67108864 -level0_file_num_compaction_trigger=4 -max_background_compactions=32 -level0_slowdown_writes_trigger=8 -level0_stop_writes_trigger=12 -num_levels=4 -min_level_to_compress=2 -compression_type=zlib -compression_ratio=0.5 -block_size=16384 -open_files=10000 -target_file_size_base=67108864 -max_bytes_for_level_base=268435456 -writes_per_range_tombstone=50000 -range_tombstone_width=50000 -max_num_range_tombstones=1000 -num=50000000./db_bench -benchmarks=readrandom -readonly -use_existing_db=true -open_files=10000 -num=50000000 -reads=1000000 -cache_size=1073741824 -threads=32