[jira] Commented: (HADOOP-6330) Integrating IBM General Parallel File System implementation of Hadoop Filesystem interface

["Steve Loughran (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HADOOP-6330?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12779937#action_12779937 ] 

Steve Loughran commented on HADOOP-6330:
----------------------------------------

It would be very tricky to integrate this with the current release process, as now you'd need a GPFS FS on the 32 bit and 64 bit machines you want to build on. It would work if the GPFS binding was a separate artifact released in sync (or shortly after) the main Hadoop releases

Someone with a GPFS cluster also needs to set up Hudson to run CI builds and tests on HDFS to catch regressions that break GPFS support sooner rather than later. Its the testing of these third-party filestores that is very weak right now.

> Integrating IBM General Parallel File System implementation of Hadoop Filesystem interface
> ------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-6330
>                 URL: https://issues.apache.org/jira/browse/HADOOP-6330
>             Project: Hadoop Common
>          Issue Type: New Feature
>          Components: fs
>         Environment: Where GPFS is supported
>            Reporter: Reshu Jain
>             Fix For: 0.18.3
>
>
> GPFS is a high performance parallel file system for GNU/Linux clusters. This patch contains the implementation of the Hadoop Filesystem Interface. There is dependency on the availability of GPFS on the host where the JNI implementation can be built.
> The patch consists of fs/gpfs classes and the JNI module in c++/libgpfs. 
> http://www-03.ibm.com/systems/clusters/software/gpfs/index.html

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