[
https://issues.apache.org/jira/browse/HADOOP-2568?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12564972#action_12564972
]
Arun C Murthy commented on HADOOP-2568:
---------------------------------------
bq. Pin reduces with consecutive IDs to nodes and have a single shuffle task
per job per node
We'll achieve similar characteristics by allocating more than one task per
heartbeat... so on a heartbeat if we assign maps {0-1} to TT1 and maps {2-3} to
TT2 ... and similarly we allocate reduces {0-1} to TT3 and reduces {2-3} to TT4
we can group map-outputs 0-3 and do a single shuffle...
This would also help cluster utilization since now we only do one task at a
time... allocating more tasks up-front seems attractive from more than one
angle.
Thoughts?
> Pin reduces with consecutive IDs to nodes and have a single shuffle task per
> job per node
> -----------------------------------------------------------------------------------------
>
> Key: HADOOP-2568
> URL: https://issues.apache.org/jira/browse/HADOOP-2568
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Reporter: Devaraj Das
> Assignee: Devaraj Das
> Fix For: 0.17.0
>
>
> The idea is to reduce disk seeks while fetching the map outputs. If we
> opportunistically pin reduces with consecutive IDs (like 5, 6, 7 ..
> max-reduce-tasks on that node) on a node, and have a single shuffle task, we
> should benefit, if for every fetch, that shuffle task fetches all the outputs
> for the reduces it is shuffling for. In the case where we have 2 reduces per
> node, we will decrease the #seeks in the map output files on the map nodes by
> 50%. Memory usage by that shuffle task would be proportional to the number of
> reduces it is shuffling for (to account for the number of ramfs instances,
> one per reduce). But overall it should help.
> Thoughts?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
