You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Greg Roelofs (JIRA)" <ji...@apache.org> on 2010/10/02 05:48:33 UTC

[jira] Commented: (MAPREDUCE-388) pipes combiner has a large memory footprint

    [ https://issues.apache.org/jira/browse/MAPREDUCE-388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12917150#action_12917150 ] 

Greg Roelofs commented on MAPREDUCE-388:
----------------------------------------

Christian, do you know if this is still an issue?  If so, I'll recommend that we try to get it into 0.22 (subject to all the other priorities, of course...).

> pipes combiner has a large memory footprint
> -------------------------------------------
>
>                 Key: MAPREDUCE-388
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-388
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>            Reporter: Christian Kunz
>
> Pipes combiner implementation can have a huge memory overhead compared to the spill size. How much, depends on the record size. E.g., an application asks for >2GB memory when io.sort.mb=500, key is 16 bytes, and value is 4 bytes.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.