hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Siddharth Seth (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (MAPREDUCE-3768) MR-2450 introduced a significant performance regression
Date Sun, 05 Feb 2012 20:33:59 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-3768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13200891#comment-13200891
] 

Siddharth Seth commented on MAPREDUCE-3768:
-------------------------------------------

The patch essentially gets rid of upto 3 seconds per task. Looking at the trace, it seems
related. OTOH, noticed that the runtime difference was 2 minutes versus 15 minutes. Unless
this hive test runs a large number of jobs / task waves - this patch probably won't fix the
issue. Have opened a separate jira for the 3 second wait - MAPREDUCE-3809.
Haven't had a chance to run the hive test. Don't have the cycles rightnow to understand how
the hive build system deals with 23 and local jars. 
                
> MR-2450 introduced a significant performance regression
> -------------------------------------------------------
>
>                 Key: MAPREDUCE-3768
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-3768
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>          Components: mrv2
>    Affects Versions: 0.23.1
>            Reporter: Eli Collins
>            Priority: Blocker
>         Attachments: stopcommunicatorpatch.txt
>
>
> MAPREDUCE-2450 introduced, or at least triggers, a significant performance regression
in Hive. With MR-2450 the execution time of TestCliDriver.skewjoin goes from 2 minutes to
15 minutes. Reverting this change from the build fixes the issue.
> Here's the relevant query:
> {noformat}
> FROM src src1 JOIN src src2 ON (src1.key = src2.key)
> INSERT OVERWRITE TABLE dest_j1 SELECT src1.key, src2.value; 
> {noformat}
> You can reproduce this by running the following from Hive 8.0 against Hadoop built from
branch-23. 
> {noformat}
> ant very-clean package test -Dtestcase=TestCliDriver -Dqfile=skewjoin.q
> {noformat}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message