[jira] [Commented] (FLINK-14214) Performance regression in TwoInputStreamOperator

["Piotr Nowojski (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-14214?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16938663#comment-16938663 ] 

Piotr Nowojski commented on FLINK-14214:
----------------------------------------

After the performance improvement of https://issues.apache.org/jira/browse/FLINK-11859, which is unrelated to this performance regression, I'm struggling to reproduce this issue. I would be afraid that trying to reproduce it, would require substantial effort, that might be even not very visible to the final user.

Additionally when serialisation optimisation (https://issues.apache.org/jira/browse/FLINK-11859) is accounted for, it looks like even with this performance regression, we have better throughput compared to 1.9.

Because of that, I would suggest to close the ticket, unless someone is willing to spend more time investigating it.

> Performance regression in TwoInputStreamOperator
> ------------------------------------------------
>
>                 Key: FLINK-14214
>                 URL: https://issues.apache.org/jira/browse/FLINK-14214
>             Project: Flink
>          Issue Type: Bug
>          Components: Runtime / Task
>    Affects Versions: 1.10.0
>            Reporter: Piotr Nowojski
>            Assignee: Piotr Nowojski
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> https://issues.apache.org/jira/browse/FLINK-13051 introduced a performance regression visible for example [here|http://codespeed.dak8s.net:8000/timeline/#/?exe=1,3&ben=twoInputMapSink&env=2&revs=200&equid=off&quarts=off&extr=off] as a drop on August 30th (note that the later performance improvement is caused by serialisation improvement, which is unrelated to the previous slow down). 
> Probable suspect is the following {{.isDone()}} check executed inside {{StreamTwoInputProcessor}} once per record:
> {code:java}
> // to avoid starvation, if the input selection is ALL and availableInputsMask is not ALL,
> // always try to check and set the availability of another input
> // TODO: because this can be a costly operation (checking volatile inside CompletableFuture`
> //  this might be optimized to only check once per processed NetworkBuffer
> if (inputSelectionHandler.shouldSetAvailableForAnotherInput()) {
> 	checkAndSetAvailable(1 - readingInputIndex);
> }
> {code}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)