You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Sagar Rao (Jira)" <ji...@apache.org> on 2021/06/08 15:50:00 UTC

[jira] [Commented] (KAFKA-12848) Add some basic benchmarks for Kafka Streams

    [ https://issues.apache.org/jira/browse/KAFKA-12848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17359432#comment-17359432 ] 

Sagar Rao commented on KAFKA-12848:
-----------------------------------

[~ableegoldman], i have added benchmarks for persistent stores for kafka streams:

[https://github.com/apache/kafka/pull/10842.|https://github.com/apache/kafka/pull/10842]

If these look fine, I will add for other kinds of state stores. This is just in continuation of the example pR by John. 

 

BTW, couple of questions:

 

1) Is there a way to run these benchmarks in a scheduled manner on the CI/CD tool used by you guys? I have seen something similar done on kubernetes. Do you think that would make sense?

2) The other thing is I need to use these rocksdb related state stores benchmarks for the couple of other tickets: Direct ByteBuffer and rocksdb merge() API. So, can we merge just this while we keep adding other jmh benchmarks or we wait for everything and i keep switching locally? Or based upon what all we want to cover in the benchmarks, we can add sub tasks? WDYT?

 

 

> Add some basic benchmarks for Kafka Streams
> -------------------------------------------
>
>                 Key: KAFKA-12848
>                 URL: https://issues.apache.org/jira/browse/KAFKA-12848
>             Project: Kafka
>          Issue Type: Improvement
>          Components: streams
>            Reporter: A. Sophie Blee-Goldman
>            Assignee: Sagar Rao
>            Priority: Major
>              Labels: newbie, newbie++
>
> As the title suggests, we often want to test out improvements or verify that a bugfix does not introduce a serious regression. While there are existing benchmarks that are run for quality assurance by various contributors, there are no publicly available benchmarks for Kafka Streams in AK itself.
> It would be great if we had a simple jmh suite (or something) with various Streams features which could be run on a one-off basis by developers.



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