[
https://issues.apache.org/jira/browse/HAMA-540?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13253303#comment-13253303
]
praveen sripati commented on HAMA-540:
--------------------------------------
1) It would be good to have the initial (p-1) pivots evenly distributed over
the input data. This way the processors take almost same time to complete the
task. How are you planning to get them?
In the TeraSort#readPartitions looks like the first reducer # of lines are read
from the input data. Not an efficient way.
http://svn.apache.org/repos/asf/hadoop/common/trunk/hadoop-mapreduce-project/hadoop-mapreduce-examples/src/main/java/org/apache/hadoop/examples/terasort/TeraSort.java
2) Slightly less than double the size of the input data is passed between the
processors in the sampling sort. If the input is 1TB, then ~2TB is transferred
between the nodes. Not sure if this is OK? It would be nice to know how much
data is transferred in merge sort and quick sort.
----
I am just getting a feel of Hama/BSP and thought of implementing sampling sort.
Please do it. I will pick the next easy one :)
> Create distributed sort BSP
> ---------------------------
>
> Key: HAMA-540
> URL: https://issues.apache.org/jira/browse/HAMA-540
> Project: Hama
> Issue Type: New Feature
> Components: bsp, examples
> Reporter: Thomas Jungblut
>
> For HAMA-535 we need some kind of sort framework, for various other tasks
> this could be as well practical.
--
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
