feat: use arrow row format for hash-group-by #4830
Merged
Conversation
This file contains 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
For apache#2723. This has two effects: - **wider feature support:** We now use the V2 aggregator for all group-column types. The arrow row format support is sufficient for that. V1 will only be used if the aggregator itself doesn't support V2 (and these are quite a few at the moment). We'll improve on that front in follow-up PRs. - **more speed:** Turns out the arrow row format is also faster (see below). Perf results (mind the noise in the benchmarks that are actually not even touched by this code change): ```text ❯ cargo bench -p datafusion --bench aggregate_query_sql -- --baseline issue2723a-pre ... Running benches/aggregate_query_sql.rs (target/release/deps/aggregate_query_sql-fdbe5671f9c3019b) aggregate_query_no_group_by 15 12 time: [779.28 µs 782.77 µs 786.28 µs] change: [+2.1375% +2.7672% +3.4171%] (p = 0.00 < 0.05) Performance has regressed. Found 1 outliers among 100 measurements (1.00%) 1 (1.00%) high mild aggregate_query_no_group_by_min_max_f64 time: [712.96 µs 715.90 µs 719.14 µs] change: [+0.8379% +1.7648% +2.6345%] (p = 0.00 < 0.05) Change within noise threshold. Found 10 outliers among 100 measurements (10.00%) 3 (3.00%) low mild 6 (6.00%) high mild 1 (1.00%) high severe Benchmarking aggregate_query_no_group_by_count_distinct_wide: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 8.7s, enable flat sampling, or reduce sample count to 50. aggregate_query_no_group_by_count_distinct_wide time: [1.7297 ms 1.7399 ms 1.7503 ms] change: [-34.647% -33.908% -33.165%] (p = 0.00 < 0.05) Performance has improved. Found 5 outliers among 100 measurements (5.00%) 5 (5.00%) high mild Benchmarking aggregate_query_no_group_by_count_distinct_narrow: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 5.7s, enable flat sampling, or reduce sample count to 60. aggregate_query_no_group_by_count_distinct_narrow time: [1.0984 ms 1.1045 ms 1.1115 ms] change: [-38.581% -38.076% -37.569%] (p = 0.00 < 0.05) Performance has improved. Found 6 outliers among 100 measurements (6.00%) 1 (1.00%) low mild 5 (5.00%) high mild Benchmarking aggregate_query_group_by: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 9.1s, enable flat sampling, or reduce sample count to 50. aggregate_query_group_by time: [1.7810 ms 1.7925 ms 1.8057 ms] change: [-25.252% -24.127% -22.737%] (p = 0.00 < 0.05) Performance has improved. Found 9 outliers among 100 measurements (9.00%) 2 (2.00%) low mild 5 (5.00%) high mild 2 (2.00%) high severe Benchmarking aggregate_query_group_by_with_filter: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. aggregate_query_group_by_with_filter time: [1.2068 ms 1.2119 ms 1.2176 ms] change: [+2.2847% +3.0857% +3.8789%] (p = 0.00 < 0.05) Performance has regressed. Found 10 outliers among 100 measurements (10.00%) 1 (1.00%) low mild 7 (7.00%) high mild 2 (2.00%) high severe Benchmarking aggregate_query_group_by_u64 15 12: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 8.7s, enable flat sampling, or reduce sample count to 50. aggregate_query_group_by_u64 15 12 time: [1.6762 ms 1.6848 ms 1.6942 ms] change: [-29.598% -28.603% -27.400%] (p = 0.00 < 0.05) Performance has improved. Found 8 outliers among 100 measurements (8.00%) 1 (1.00%) low mild 1 (1.00%) high mild 6 (6.00%) high severe Benchmarking aggregate_query_group_by_with_filter_u64 15 12: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. aggregate_query_group_by_with_filter_u64 15 12 time: [1.1969 ms 1.2008 ms 1.2049 ms] change: [+1.3015% +2.1513% +3.0016%] (p = 0.00 < 0.05) Performance has regressed. Found 6 outliers among 100 measurements (6.00%) 1 (1.00%) low severe 2 (2.00%) high mild 3 (3.00%) high severe aggregate_query_group_by_u64_multiple_keys time: [14.797 ms 15.112 ms 15.427 ms] change: [-12.072% -8.7274% -5.3392%] (p = 0.00 < 0.05) Performance has improved. Found 3 outliers among 100 measurements (3.00%) 3 (3.00%) high mild aggregate_query_approx_percentile_cont_on_u64 time: [4.1278 ms 4.1687 ms 4.2098 ms] change: [+0.4851% +1.9525% +3.3676%] (p = 0.01 < 0.05) Change within noise threshold. Found 2 outliers among 100 measurements (2.00%) 1 (1.00%) low mild 1 (1.00%) high mild aggregate_query_approx_percentile_cont_on_f32 time: [3.4694 ms 3.4967 ms 3.5245 ms] change: [-1.5467% -0.4432% +0.6609%] (p = 0.43 > 0.05) No change in performance detected. Found 1 outliers among 100 measurements (1.00%) 1 (1.00%) high mild ```
github-actions
bot
added
core
tustvold
approved these changes
Jan 6, 2023
alamb
approved these changes
Jan 6, 2023
There was a problem hiding this comment.
wider data format support and faster. What is not to love 👍
Thank you @crepererum
| @@ -90,7 +90,7 @@ struct GroupedHashAggregateStreamV2Inner { | |||
| group_by: PhysicalGroupBy, | |||
| accumulators: Vec<AccumulatorItemV2>, | |||
|
|
|||
| group_schema: SchemaRef, | |||
There was a problem hiding this comment.
🤔 I wonder how far we are away from being able to remove datafusion_row 🤔
|
Benchmark runs are scheduled for baseline = 169b522 and contender = bc6a2dc. bc6a2dc is a master commit associated with this PR. Results will be available as each benchmark for each run completes. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Which issue does this PR close?
For #2723.
Rationale for this change
This has two effects:
What changes are included in this PR?
Use arrow row format for group keys instead of DF row format.
Are these changes tested?
Perf results (mind the noise in the benchmarks that are actually not even touched by this code change):
Are there any user-facing changes?
Faster group-bys.