[ 
https://issues.apache.org/jira/browse/HAMA-535?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13272606#comment-13272606
 ] 

Suraj Menon edited comment on HAMA-535 at 5/10/12 6:30 PM:
-----------------------------------------------------------

Hi, Please check the link below for an outline of the solution that I want to 
propose. The key difference here from the Hadoop MR is that since BSP Task can 
communicate with each other, it can share the existent key distribution among 
each other. We can have different strategies to choose which reducer to run on 
what set of keys based on the key distribution that every peer gets. This is 
handled in ReducerKeyDesignator.java.

https://github.com/ssmenon/hama/tree/master/hama-mapreduce/src/org/apache/hama/computemodel/mapreduce

Disclaimer: The code is incomplete and not tested. The code is written only to 
outline the solution.
I will be testing it for small set of data, before we get into spilling and 
merging.
                
      was (Author: surajsmenon):
    Hi, Please check the link below for an outline of the solution that I want 
to propose. The key difference here from the Hadoop MR is that since BSP Task 
can communicate with each other, it can share the existent key distribution 
among each other. We can have different strategies to choose which reducer to 
run on what set of keys based on the key distribution that every peer gets. 
This is handled in ReducerKeyDesignator.java.

https://github.com/ssmenon/hama/tree/master/hama-mapreduce/src/org/apache/hama/computemodel/mapreduce
                  
> Implement MapReduce in BSP
> --------------------------
>
>                 Key: HAMA-535
>                 URL: https://issues.apache.org/jira/browse/HAMA-535
>             Project: Hama
>          Issue Type: New Feature
>          Components: examples
>            Reporter: Thomas Jungblut
>            Assignee: Suraj Menon
>
> Suraj had the idea of implementing MapReduce on top of Apache Hama.
> This is generally possible because BSP is a generalization to MapReduce.
> We should just make up a simple API compatible version to Hadoop and put it 
> into our example module.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to