[
https://issues.apache.org/jira/browse/MAPREDUCE-4049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13567748#comment-13567748
]
Avner BenHanoch commented on MAPREDUCE-4049:
--------------------------------------------
Hi Alejadro & Arun,
Thank you for your review and all your comments. I appreciate your help and
responsiveness with my issue.
I would like to say a few comments/answers before the patch is concluded:
1. *_Alejandro_* - _getJobConf(JobID)_ is needed for any ShuffleProvider. The
provider needs it for determining _username_ and _runAsUsername_. _username_
is needed for determining the location in disk of the MOF and Index files.
_runAsUsername_ is needed for reading the above files with the right privileges.
2. *_Alejandro_* – The answer for your question about the tests is - YES. I did
run all smoke & commit tests successfully.
3. *_Arun_* - I have no problem with your request for not passing the entire
ReduceTask. I am only a bit worried about initing ShuffleConsumerPlugin with
arguments such as _getPartition()_ and _getJobTokenSecret()_. The reason is
that at least theoretically it is possible to change _partition/jobTokenSecret_
after the shuffleConsumerPlugin was initiated. Hence, I need your approval for
that.
Additionally, please notice that in hadoop-trunk we do pass the entire
ReduceTask to the ShuffleConsumerPlugin. (Also, in hadoop-1 we always passed
ReduceTask. I think that with the last patch it is highlighted because we made
ReduceCopier a static class which required specifying explicitly reduceTask.XXX
in about 75 different places).
*_Bottom line, Arun, please let me know if you are still worried about passing
the entire ReduceTask to the shuffle plugin._*
thank you,
Avner
> plugin for generic shuffle service
> ----------------------------------
>
> Key: MAPREDUCE-4049
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-4049
> Project: Hadoop Map/Reduce
> Issue Type: Sub-task
> Components: performance, task, tasktracker
> Affects Versions: 1.0.3, 1.1.0, 2.0.0-alpha, 3.0.0
> Reporter: Avner BenHanoch
> Assignee: Avner BenHanoch
> Labels: merge, plugin, rdma, shuffle
> Fix For: 2.0.3-alpha
>
> Attachments: HADOOP-1.x.y.patch, Hadoop Shuffle Plugin Design.rtf,
> MAPREDUCE-4049--branch-1.patch, MAPREDUCE-4049--branch-1.patch,
> mapreduce-4049.patch
>
>
> Support generic shuffle service as set of two plugins: ShuffleProvider &
> ShuffleConsumer.
> This will satisfy the following needs:
> # Better shuffle and merge performance. For example: we are working on
> shuffle plugin that performs shuffle over RDMA in fast networks (10gE, 40gE,
> or Infiniband) instead of using the current HTTP shuffle. Based on the fast
> RDMA shuffle, the plugin can also utilize a suitable merge approach during
> the intermediate merges. Hence, getting much better performance.
> # Satisfy MAPREDUCE-3060 - generic shuffle service for avoiding hidden
> dependency of NodeManager with a specific version of mapreduce shuffle
> (currently targeted to 0.24.0).
> References:
> # Hadoop Acceleration through Network Levitated Merging, by Prof. Weikuan Yu
> from Auburn University with others,
> [http://pasl.eng.auburn.edu/pubs/sc11-netlev.pdf]
> # I am attaching 2 documents with suggested Top Level Design for both plugins
> (currently, based on 1.0 branch)
> # I am providing link for downloading UDA - Mellanox's open source plugin
> that implements generic shuffle service using RDMA and levitated merge.
> Note: At this phase, the code is in C++ through JNI and you should consider
> it as beta only. Still, it can serve anyone that wants to implement or
> contribute to levitated merge. (Please be advised that levitated merge is
> mostly suit in very fast networks) -
> [http://www.mellanox.com/content/pages.php?pg=products_dyn&product_family=144&menu_section=69]
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
