You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@kafka.apache.org by "Edward Ribeiro (JIRA)" <ji...@apache.org> on 2015/09/01 21:56:46 UTC

[jira] [Commented] (KAFKA-2499) kafka-producer-perf-test should use something more realistic than empty byte arrays

    [ https://issues.apache.org/jira/browse/KAFKA-2499?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14726069#comment-14726069 ] 

Edward Ribeiro commented on KAFKA-2499:
---------------------------------------

Hi [~benstopford], I have a tidy bit of previous experience with synthetic data generation. If you are not going to work on this, I can provide some additional code if you assign this issue to me. Or I can provide you some classes for generating those random values. Up to you. :)

> kafka-producer-perf-test should use something more realistic than empty byte arrays
> -----------------------------------------------------------------------------------
>
>                 Key: KAFKA-2499
>                 URL: https://issues.apache.org/jira/browse/KAFKA-2499
>             Project: Kafka
>          Issue Type: Bug
>            Reporter: Ben Stopford
>
> ProducerPerformance.scala (There are two of these, one used by the shell script and one used by the system tests. Both exhibit this problem)
> creates messags from empty byte arrays. 
> This is likely to provide unrealistically fast compression and hence unrealistically fast results. 
> Suggest randomised bytes or more realistic sample messages are used. 
> Thanks to Prabhjot Bharaj for reporting this. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)