hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jothi Padmanabhan (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, 30 Jan 2009 09:47:59 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-1338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12668794#action_12668794
] 

Jothi Padmanabhan commented on HADOOP-1338:
-------------------------------------------

No, there is no impact on batching 10 outputs at a time on gridmix runs. An appreciable difference
in shuffle time was observed with the loadgen 100 byte, 100000 maps on a 100 node cluster
with 1 reducer (Trunk - 1m24s, Patch - 52s). However, for these kind of applications, the
shuffle time is fairly negligible compared to the map run time (50 mins or so). 

> 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
>            Assignee: Jothi Padmanabhan
>
> 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