[jira] Commented: (HADOOP-3097) mapred.map.tasks and mapred.reduce.tasks should be determined automatically by default

["Owen O'Malley (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HADOOP-3097?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12587902#action_12587902 ] 

Owen O'Malley commented on HADOOP-3097:
---------------------------------------

+1 for using the cluster defaults for this. I might suggest using #map slots * 2.0 and # reduce slots * 0.99 since those seem to be a little better defaults that 1.0 for both maps and reduces.

> mapred.map.tasks and mapred.reduce.tasks should be determined automatically by default
> --------------------------------------------------------------------------------------
>
>                 Key: HADOOP-3097
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3097
>             Project: Hadoop Core
>          Issue Type: New Feature
>          Components: mapred
>            Reporter: Doug Cutting
>
> The default values for mapred.map.tasks and mapred.reduce.tasks should be empty or -1, signalling that the framework should use an dynamically-determined default.  An appropriate default is perhaps the number of map and reduce slots in the cluster, since FileInputFormat interprets mapred.map.tasks as a minimum.
> This would remove a common area of misconfiguration.

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