Mailing-List: contact mapreduce-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: mapreduce-issues@hadoop.apache.org
Date: Wed, 14 Sep 2011 13:13:09 +0000 (UTC)
From: "Amar Kamat (JIRA)" <jira@apache.org>
To: mapreduce-issues@hadoop.apache.org
Message-ID: 
 <791575839.26553.1316005989431.JavaMail.tomcat@hel.zones.apache.org>
In-Reply-To: 
 <1597268113.26544.1316005755050.JavaMail.tomcat@hel.zones.apache.org>
Subject: [jira] [Updated] (MAPREDUCE-3008) [Gridmix] Improve cumulative CPU
 usage emulation for short running tasks
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/MAPREDUCE-3008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Amar Kamat updated MAPREDUCE-3008:
----------------------------------

    Attachment: mapreduce-2591-v1.4.2.patch

Attaching a patch that improves CPU emulation for short running tasks. Areas of improvements:
1. Sorter/Comparator now is CPU emulation aware
2. For tasks with no spills/merges, aggressive CPU emulation is done.

tets-patch and JUnit tests for Gridmix passed.

> [Gridmix] Improve cumulative CPU usage emulation for short running tasks
> ------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-3008
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-3008
>             Project: Hadoop Map/Reduce
>          Issue Type: Sub-task
>          Components: contrib/gridmix
>    Affects Versions: 0.24.0
>            Reporter: Amar Kamat
>              Labels: cpu-emulation, gridmix
>             Fix For: 0.24.0
>
>         Attachments: mapreduce-2591-v1.4.2.patch
>
>
> CPU emulation in Gridmix fails to meet the expected target if the map has no data to sort/spill/merge. There are 2 major reasons for this:
> 1. The map task end immediately ends soon after the map task. The map progress is 67% while the map phase ends. 
> 2. Currently, the sort (comparator) doesnt emulate CPU. If the map is short lived, the CPU emulation thread (spawned from the map task in cleanup) doesn't get a chance to emulate.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira