[
https://issues.apache.org/jira/browse/HADOOP-3297?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12592842#action_12592842
]
Runping Qi commented on HADOOP-3297:
------------------------------------
Under what condition fetched map outputs will end up on disk directly?
If a segment is very large, it makes sense to write it out on disk directly.
If it is one or the last few, it makes sense too. Otherwise, a fetched segment
should get into in-mem file system. If the in-mem file is full, the fetcher
should wait.
This is related to hadoop-2095. They should be considered together.
> The way in which ReduceTask/TaskTracker gets completion events during shuffle
> can be improved
> ---------------------------------------------------------------------------------------------
>
> Key: HADOOP-3297
> URL: https://issues.apache.org/jira/browse/HADOOP-3297
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Reporter: Devaraj Das
> Assignee: Devaraj Das
> Fix For: 0.18.0
>
>
> Certain things like poll frequency, number of events fetched in one go, etc.
> can probably be improved to improve the shuffle performance. This would
> affect the task-->tasktracker and the tasktracker-->jobtracker shuffle
> related RPCs.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
