hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Raghu Angadi (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 Wed, 19 Nov 2008 23:21:44 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-1338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12649209#action_12649209

Raghu Angadi commented on HADOOP-1338:

I think on a LAN, overhead of a establishing a TCP connection usually gets overrated. Slow
start is a real problem. Just like how this could be disabled in Hadoop, Jetty might have
an option. 

I would think just disabling slow-start would give more benefit than keep-alive without disabling

In any case, already existing benchmarks might be able to show any benefit  this gives. Given
possible side affects of keep-alive as Devaraj mentioned, improvement might need to be reasonably

> 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.

View raw message