[ 
https://issues.apache.org/jira/browse/MAPREDUCE-279?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13000464#comment-13000464
 ] 

Arun C Murthy commented on MAPREDUCE-279:
-----------------------------------------

@Min 

bq. How does ApplicationMaster know its resource requirements before it 
launches tasks?
The assumption is that the AM has a basic idea about resource requirements for 
it's application which is feasible for our primary use case: Map-Reduce. OTOH, 
an AM for other applications has the ability to launch a few tasks, watch their 
resource consumption/utilization and update future resource requests.

bq. Even common users can deploy their ApplicationMaster over the cluster they 
have no any permissions on that?
>From the framework (i.e. RM/NN) perspective, everything in the cluster 
>including AMs is 'user-land'. Thus as long as a user implements the protocols 
>for AMs they can deploy any applications... they do not need any permission to 
>deploy.

I'm working very hard to get the codebase committed to a branch, once there we 
would love your f/b on the protocols etc. Hopefully that should help you 
understand how to implement a custom AM if you so choose... appreciate your 
patience while I work the system! *smile*


> Map-Reduce 2.0
> --------------
>
>                 Key: MAPREDUCE-279
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-279
>             Project: Hadoop Map/Reduce
>          Issue Type: Improvement
>          Components: jobtracker, tasktracker
>            Reporter: Arun C Murthy
>            Assignee: Arun C Murthy
>             Fix For: 0.23.0
>
>
> Re-factor MapReduce into a generic resource scheduler and a per-job, 
> user-defined component that manages the application execution. 

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

        

Reply via email to