You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "tartarus (Jira)" <ji...@apache.org> on 2020/09/10 02:50:00 UTC
[jira] [Commented] (FLINK-17137) Support mini batch for
WindowOperator in blink planner
[ https://issues.apache.org/jira/browse/FLINK-17137?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17193316#comment-17193316 ]
tartarus commented on FLINK-17137:
----------------------------------
[~libenchao] hello, Is this to reduce state access to improve performance? It seems that the benefits will be more obvious under rocksdb.
> Support mini batch for WindowOperator in blink planner
> ------------------------------------------------------
>
> Key: FLINK-17137
> URL: https://issues.apache.org/jira/browse/FLINK-17137
> Project: Flink
> Issue Type: Improvement
> Components: Table SQL / Planner, Table SQL / Runtime
> Reporter: Benchao Li
> Priority: Major
>
> Currently only regular aggregate and deduplicate support mini batch. WindowOperator is a very frequently used operator in Flink, it's very helpful to support mini batch for it.
> Design document: https://docs.google.com/document/d/1GYlrg8dkYcw5fuq1HptdA3lygXrS_VRbnI8NXoEtZCg/edit?usp=sharing
> cc [~jark]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)