hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Runping Qi (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-1338) Improve the shuffle phase by using the "connection: keep-alive" and doing batch transfers of files
Date Fri, 27 Jun 2008 18:03:45 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-1338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12608868#action_12608868
] 

Runping Qi commented on HADOOP-1338:
------------------------------------


Batch transfer may not be a big deal before 0.18, because there were a lot of inefficient
places in the shuffling code anyway.
However, it will be a significant improvement for 0.18.
Based on my observations on running gridmix with 0.18, the fatching part becomes a bottleneck
of the shuffling phase in 0.18.
I think it is definitely worth to revisit this issue.
Based on my understanding of the shuffling code, it is not hard to implement keep alive connection
in the fetcher.



> Improve the shuffle phase by using the "connection: keep-alive" and doing batch transfers
of files
> --------------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-1338
>                 URL: https://issues.apache.org/jira/browse/HADOOP-1338
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: mapred
>            Reporter: Devaraj Das
>
> We should do transfers of map outputs at the granularity of  *total-bytes-transferred*
rather than the current way of transferring a single file and then closing the connection
to the server. A single TaskTracker might have a couple of map output files for a given reduce,
and we should transfer multiple of them (upto a certain total size) in a single connection
to the TaskTracker. Using HTTP-1.1's keep-alive connection would help since it would keep
the connection open for more than one file transfer. We should limit the transfers to a certain
size so that we don't hold up a jetty thread indefinitely (and cause timeouts for other clients).
> Overall, this should give us improved performance.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message