Multi get #226
Open
Multi get #226
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
biryukovmaxim
force-pushed
the
multi-get
branch
from
42ab8de to
0708f12
Compare
biryukovmaxim
force-pushed
the
multi-get
branch
from
3ee367f to
80b6e78
Compare
…ng and improve domain/op structure.
…ser_data` for cleaner user data handling
…g for too long key set
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR introduces a parallel implementation of multi_get using io_uring for concurrent I/O operations, which significantly reduces latency in scenarios where keys are scattered across many SST files and involve large, I/O-bound reads (e.g., from cold page cache or slow storage)
Benchmark Results
A new ignored benchmark test (multi_get_scattered_large_values_outperforms_single_gets) validates this:
Setup: 64 keys scattered across 64 SSTs, 1 MiB values, compression/filters/block cache disabled, page cache dropped before each run.
Results (on my machine, NVMe SSD):
Average sequential gets time: 67.76 ms
Average multi_get time: 27.39 ms
Speedup: ~2.5x
How to Run the Benchmark
This test requires root privileges (for dropping caches via sudo) and is ignored by default. Run it in release mode for accurate perf:
cargo test --color=always --package lsm-tree --test tree_multi_get sudo_required::multi_get_scattered_large_values_outperforms_single_gets --profile release -- --nocapture --ignoredPrefix with sudo if needed for cache dropping. Adjust num_ssts or value_size in the test for experimentation.
Notes
The test uses sudo in drop_caches() to ensure cold I/O; without it, page cache masks the benefit.