You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Siddharth Seth (JIRA)" <ji...@apache.org> on 2015/06/05 00:37:37 UTC
[jira] [Commented] (TEZ-2473) Consider using RawLocalFileSystem in
MapOutput.createDiskMapOutput
[ https://issues.apache.org/jira/browse/TEZ-2473?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14573709#comment-14573709 ]
Siddharth Seth commented on TEZ-2473:
-------------------------------------
[~rajesh.balamohan] - we'll need to do this for DIskFetchedInput as well.
Should this changed be linked to increasing the frequency at which ifile writes checksums ? I believe that's a 4 byte sum irresopective of the data size ? The output on the 'map' side is already using RawFs - so is relying on the ifile checksum. I think it should be ok to change this as well, independent of the Ifile checksum, but that would be good to fix sooner rather than later.
> Consider using RawLocalFileSystem in MapOutput.createDiskMapOutput
> ------------------------------------------------------------------
>
> Key: TEZ-2473
> URL: https://issues.apache.org/jira/browse/TEZ-2473
> Project: Apache Tez
> Issue Type: Bug
> Reporter: Rajesh Balamohan
> Assignee: Rajesh Balamohan
> Attachments: TEZ-2473.1.patch
>
>
> Currently it makes use of LocalFileSystem which would go through checksumming. This can save some CPU cycles in tasks involving disk merges.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)