You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-dev@hadoop.apache.org by "Devaraj Das (JIRA)" <ji...@apache.org> on 2008/05/22 09:03:55 UTC

[jira] Updated: (HADOOP-3427) In ReduceTask::fetchOutputs, wait for result can be improved slightly

     [ https://issues.apache.org/jira/browse/HADOOP-3427?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Devaraj Das updated HADOOP-3427:
--------------------------------

    Attachment: 3427.patch

This patch mainly introduces this change - Defines a number to indicate when the shuffle scheduler should stop and just wait for notifications from the fetchers. This number is hardcoded to (4 * numCopiers). 

> In ReduceTask::fetchOutputs, wait for result can be improved slightly
> ---------------------------------------------------------------------
>
>                 Key: HADOOP-3427
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3427
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: mapred
>            Reporter: Devaraj Das
>            Assignee: Devaraj Das
>             Fix For: 0.18.0
>
>         Attachments: 3427.patch
>
>
> The getCopyResult call in ReduceTask::fetchOutputs waits for 2 seconds if the results list is empty. This can be improved to wait only when sufficient number of fetches (above a certain threshold) have been scheduled. The threshold should ensure that all the fetcher threads would be busy enough, and in this case, the call to getCopyResult would return only when a fetcher thread wakes it up.

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