[
https://issues.apache.org/jira/browse/BEAM-9327?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Etienne Chauchot resolved BEAM-9327.
------------------------------------
Fix Version/s: Not applicable
Resolution: Won't Fix
recoding adding an input to an accumulator translation with manual window
merging is less performant (mesured with nexmark) than calling
mergeAccumulators(accumulator, addInput(createAccumulator, input))
Using current SparkCombineFn in the new runner is less performant also.
> Avoid creating a new accumulator for each add input in Combine translation
> --------------------------------------------------------------------------
>
> Key: BEAM-9327
> URL: https://issues.apache.org/jira/browse/BEAM-9327
> Project: Beam
> Issue Type: Improvement
> Components: runner-spark
> Reporter: Etienne Chauchot
> Assignee: Etienne Chauchot
> Priority: Major
> Labels: structured-streaming
> Fix For: Not applicable
>
>
> similar to latest inprovement in the current runner. See:
> [https://www.youtube.com/watch?v=ZIFtmx8nBow&t=721s] min 12
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
