[
https://issues.apache.org/jira/browse/FLINK-25750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17480036#comment-17480036
]
Roman Khachatryan commented on FLINK-25750:
-------------------------------------------
Thanks for the confirmation and sharing your thoughts [~dmvk] and [~chesnay].
I agree, excluding job submission from execution time would eliminate some of
the regressions.
One potential difficulty I see is actualy starting the execution. It can be
some form of source synchronization, but it needs to be customizable for
different graphs (or maybe all benchmark graphs are simple enough?)
cc: [~pnowojski]
> Performance regression on 20.01.2021 in globalWindow and stateBackend
> benchmarks
> --------------------------------------------------------------------------------
>
> Key: FLINK-25750
> URL: https://issues.apache.org/jira/browse/FLINK-25750
> Project: Flink
> Issue Type: Bug
> Components: Benchmarks, Runtime / State Backends
> Affects Versions: 1.15.0
> Reporter: Roman Khachatryan
> Assignee: Roman Khachatryan
> Priority: Critical
> Fix For: 1.15.0
>
>
> http://codespeed.dak8s.net:8000/timeline/#/?exe=1,3&ben=globalWindow&env=2&revs=200&equid=off&quarts=on&extr=on
> http://codespeed.dak8s.net:8000/timeline/#/?exe=1,3&ben=stateBackends.FS&env=2&revs=200&equid=off&quarts=on&extr=on
> http://codespeed.dak8s.net:8000/timeline/#/?exe=1,3&ben=stateBackends.FS&env=2&revs=200&equid=off&quarts=on&extr=on
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
