[
https://issues.apache.org/jira/browse/BEAM-9748?focusedWorklogId=422811&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-422811
]
ASF GitHub Bot logged work on BEAM-9748:
----------------------------------------
Author: ASF GitHub Bot
Created on: 15/Apr/20 16:09
Start Date: 15/Apr/20 16:09
Worklog Time Spent: 10m
Work Description: lukecwik commented on pull request #11406: [BEAM-9748]
Move Reparallelize transform to Reshuffle
URL: https://github.com/apache/beam/pull/11406#discussion_r408960107
##########
File path:
sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/Reshuffle.java
##########
@@ -65,6 +66,16 @@ private Reshuffle() {}
return new ViaRandomKey<>();
}
+ @Experimental
Review comment:
I think the key is that the large input was generated with limited
parallelism and suggest that we update `ViaRandomKey` to perform both
expansions where the side input fusion break is added if the
`withHintHighFanoutAndLimitedInputParallelism()` is specified.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 422811)
Time Spent: 2h 40m (was: 2.5h)
> Add Reshuffle.ForSequentiallyGeneratedInput transform
> -----------------------------------------------------
>
> Key: BEAM-9748
> URL: https://issues.apache.org/jira/browse/BEAM-9748
> Project: Beam
> Issue Type: Improvement
> Components: sdk-java-core
> Reporter: Ismaël Mejía
> Assignee: Ismaël Mejía
> Priority: Minor
> Time Spent: 2h 40m
> Remaining Estimate: 0h
>
> Some DoFn based IOs like JdbcIO and RedisIO rely on a different approach to
> Reparallelize outputs using a combination of a an empty PCollectionView to
> force materialization and Reshuffle.viaRandomkey to reparallelize a
> PCollection. This issue extracts this transform and expose it as part of the
> Reshuffle to avoid repeating the code for transforms (notably IOs) that
> produce lots of sequentially generated data where and benefit of this
> alternative approach to perform better reparallelization of its output.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
