[
https://issues.apache.org/jira/browse/FLINK-23308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17389747#comment-17389747
]
Dawid Wysakowicz commented on FLINK-23308:
------------------------------------------
After running benchmarks on the aforementioned commit range I found out that
e.g.
(https://github.com/apache/flink/commit/b094a932845db5539fc07b032d49d0bcefd15df2)
shows some performance regression which we can not explain. It does not change
any classes that are used in the readFileSplit.
Notice we could not reproduce the 34k records for any runs, even for the same
flink and benchmarks commits that did so some time in the past.
The last "good" commit that does not show a regression, and reaches 33,5k was
8bb629460ae. After that the performance falls down a bit to ~32k. However, the
confidence interval 99.9% is in the range ~31k-34k. Taking that into account
let's assume we cannot reproduce the behaviour. We might revisit it if shows up
in some other benchmarks.
> Performance regression on 06.07
> -------------------------------
>
> Key: FLINK-23308
> URL: https://issues.apache.org/jira/browse/FLINK-23308
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Task
> Affects Versions: 1.14.0
> Reporter: Dawid Wysakowicz
> Assignee: Dawid Wysakowicz
> Priority: Critical
> Labels: pull-request-available
> Fix For: 1.14.0
>
>
> http://codespeed.dak8s.net:8000/timeline/#/?exe=1,3&ben=twoInputMapSink&env=2&revs=200&equid=off&quarts=on&extr=on
> http://codespeed.dak8s.net:8000/timeline/?ben=readFileSplit&env=2
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
