[
https://issues.apache.org/jira/browse/FLINK-27556?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17539482#comment-17539482
]
Anton Kalashnikov commented on FLINK-27556:
-------------------------------------------
[~roman], as I can see the cause of this regression is FLINK-25511(the first
commit 5b58ef66a8835bf22db23e1d9836a1e9f4d94045). More precisely, the reason is
a new field in ByteStreamStateHandle. As I understand, our microbenchmark is
too sensitive to even such a small extra data which we write to each checkpoint
now. So we need to think about how it is important for us.
Btw, right now we write 'handleName' twice - directly in
'ByteStreamStateHandle' and inside 'physicalID'. Was it the target? or we can
avoid duplication?
> Performance regression in checkpointSingleInput.UNALIGNED on 29.04.2022
> -----------------------------------------------------------------------
>
> Key: FLINK-27556
> URL: https://issues.apache.org/jira/browse/FLINK-27556
> Project: Flink
> Issue Type: Bug
> Components: Benchmarks, Runtime / Checkpointing
> Affects Versions: 1.16.0
> Reporter: Roman Khachatryan
> Priority: Major
> Attachments: Screenshot_2022-05-09_10-35-57.png
>
>
> http://codespeed.dak8s.net:8000/timeline/#/?exe=1&ben=checkpointSingleInput.UNALIGNED&extr=on&quarts=on&equid=off&env=2&revs=200
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
