clintropolis commented on issue #6066: Sorting rows when rollup is disabled URL: https://github.com/apache/incubator-druid/issues/6066#issuecomment-410183701 Ok, so I had to know, so I went ahead and did benchmarks if we do the other way and sort at persist time. no rollup opportunity: ``` Benchmark (rollup) (rowsPerSegment) (schema) Mode Cnt Score Error Units IndexPersistBenchmark.persistV9 true 75000 basic avgt 25 499315.212 ± 154036.971 us/op IndexPersistBenchmark.persistV9 false 75000 basic avgt 25 449792.742 ± 28218.504 us/op IndexPersistBenchmark.persistV9 false (ordered) 75000 basic avgt 25 508051.563 ± 63033.662 us/op all size: [3038874] bytes. ``` moderate rollup opportunity: ``` Benchmark (rollup) (rowsPerSegment) (schema) Mode Cnt Score Error Units IndexPersistBenchmark.persistV9 true 75000 basic avgt 25 406840.576 ± 20732.769 us/op IndexPersistBenchmark.persistV9 false 75000 basic avgt 25 431725.214 ± 18793.693 us/op IndexPersistBenchmark.persistV9 false (ordered) 75000 basic avgt 25 494056.572 ± 34396.770 us/op rollup: size [2285574] bytes. no-rollup: size [2741399] bytes. ordered-no-rollup: size [2516639] bytes. ``` more rollup: ``` Benchmark (rollup) (rowsPerSegment) (schema) Mode Cnt Score Error Units IndexPersistBenchmark.persistV9 true 75000 basic avgt 25 338251.339 ± 22031.319 us/op IndexPersistBenchmark.persistV9 false 75000 basic avgt 25 443272.327 ± 25099.425 us/op IndexPersistBenchmark.persistV9 false (ordered) 75000 basic avgt 25 552234.263 ± 41889.207 us/op rollup: size [1755456] bytes. no-rollup: size [2741017] bytes. ordered-no-rollup: size [2346649] bytes. ``` I don't have strong feelings about the best way to do this, persist performance cost looks to be on the range of 15-20% slower here. Maybe better to sort at persist time to not risk impact to query performance?
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org For additional commands, e-mail: commits-h...@druid.apache.org