[SPARK-48518][CORE] Make LZF compression be able to run in parallel#46858
Closed
yaooqinn wants to merge 3 commits intoapache:masterfrom
Closed
[SPARK-48518][CORE] Make LZF compression be able to run in parallel#46858yaooqinn wants to merge 3 commits intoapache:masterfrom
yaooqinn wants to merge 3 commits intoapache:masterfrom
Conversation
yaooqinn
commented
Jun 4, 2024
| AMD EPYC 7763 64-Core Processor | ||
| Compress large objects: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative | ||
| ----------------------------------------------------------------------------------------------------------------------------- | ||
| Compression 1024 array values in 1 threads 39 45 5 0.0 38475.4 1.0X |
Member
Author
There was a problem hiding this comment.
With GitHub standard action runners, it seems that we only get 1 thread.
Contributor
There was a problem hiding this comment.
hmm...
[info] OpenJDK 64-Bit Server VM 17.0.10+7-LTS on Mac OS X 14.5
[info] Apple M2 Max
[info] Compress small objects: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
[info] --------------------------------------------------------------------------------------------------------------------------------
[info] Compression 256000000 int values in parallel 548 550 2 467.0 2.1 1.0X
[info] Compression 256000000 int values single-threaded 522 523 1 490.5 2.0 1.1X
[info] Running benchmark: Compress large objects
[info] Running case: Compression 1024 array values in 8 threads
[info] Stopped after 123 iterations, 2009 ms
[info] Running case: Compression 1024 array values single-threaded
[info] Stopped after 83 iterations, 2003 ms
[info] OpenJDK 64-Bit Server VM 17.0.10+7-LTS on Mac OS X 14.5
[info] Apple M2 Max
[info] Compress large objects: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
[info] -----------------------------------------------------------------------------------------------------------------------------
[info] Compression 1024 array values in 8 threads 12 16 13 0.1 11546.1 1.0X
[info] Compression 1024 array values single-threaded 23 24 1 0.0 22767.9 0.5X
I ran this benchmark locally, and it seems that the performance of Compression 256000000 int values single-threaded and Compression 256000000 int values in parallel is almost the same
Member
Author
There was a problem hiding this comment.
I guess the rate is limited by the producer
LuciferYang
reviewed
Jun 4, 2024
| .intConf | ||
| .createWithDefault(1) | ||
|
|
||
| private[spark] val IO_COMPRESSION_LZF_PARALLEL = |
Contributor
There was a problem hiding this comment.
Should we add an explanation of this configuration in configuration.md
Member
Author
|
Thank you @LuciferYang Merged to master |
mridulm
reviewed
Jun 4, 2024
| override def compressedOutputStream(s: OutputStream): OutputStream = { | ||
| new LZFOutputStream(s).setFinishBlockOnFlush(true) | ||
| if (parallelCompression) { | ||
| new PLZFOutputStream(s) |
Contributor
There was a problem hiding this comment.
This is creating a threadpool per compressedOutputStream - which can end up being quite expensive (num thread is num processors + some 'interesting' logic which tries to modulate it) .
Did you get a chance to try this on some nontrivial jobs ? Very curious about the experience.
Given this is turned off by default, dont see any concerns with the change itself though ! Would be a good way to understand the impact.
dongjoon-hyun
added a commit
that referenced
this pull request
Oct 14, 2025
…` by default ### What changes were proposed in this pull request? This PR aims to enable `spark.io.compression.lzf.parallel.enabled` by default at Apache Spark 4.1.0. ### Why are the changes needed? `spark.io.compression.lzf.parallel.enabled` was introduced at Apache Spark 4.0.0 and has been used stably so far. We can enable this by default. - #46858 ### Does this PR introduce _any_ user-facing change? Yes for `LZF` users. The migration guide is updated. ### How was this patch tested? Pass the CIs. ### Was this patch authored or co-authored using generative AI tooling? No. Closes #52603 from dongjoon-hyun/SPARK-53896. Authored-by: Dongjoon Hyun <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
huangxiaopingRD
pushed a commit
to huangxiaopingRD/spark
that referenced
this pull request
Nov 25, 2025
…` by default ### What changes were proposed in this pull request? This PR aims to enable `spark.io.compression.lzf.parallel.enabled` by default at Apache Spark 4.1.0. ### Why are the changes needed? `spark.io.compression.lzf.parallel.enabled` was introduced at Apache Spark 4.0.0 and has been used stably so far. We can enable this by default. - apache#46858 ### Does this PR introduce _any_ user-facing change? Yes for `LZF` users. The migration guide is updated. ### How was this patch tested? Pass the CIs. ### Was this patch authored or co-authored using generative AI tooling? No. Closes apache#52603 from dongjoon-hyun/SPARK-53896. Authored-by: Dongjoon Hyun <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
What changes were proposed in this pull request?
This PR introduced a config that turns on LZF compression to parallel mode via using PLZFOutputStream.
FYI, https://github.com/ning/compress?tab=readme-ov-file#parallel-processing
Why are the changes needed?
Improve performance
Does this PR introduce any user-facing change?
no
How was this patch tested?
benchmark
Was this patch authored or co-authored using generative AI tooling?
no