[jira] Updated: (HADOOP-1721) Rack-aware Shuffle

["Arun C Murthy (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HADOOP-1721?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Arun C Murthy updated HADOOP-1721:
----------------------------------

        Fix Version/s: 0.15.0
    Affects Version/s: 0.14.2
               Status: Patch Available  (was: Open)

> Rack-aware Shuffle
> ------------------
>
>                 Key: HADOOP-1721
>                 URL: https://issues.apache.org/jira/browse/HADOOP-1721
>             Project: Hadoop
>          Issue Type: Improvement
>          Components: mapred
>    Affects Versions: 0.14.2
>            Reporter: Arun C Murthy
>            Assignee: Arun C Murthy
>             Fix For: 0.15.0
>
>
> We could try and experiment with *rack-aware* scheduling of fetches per-reducer. Given the disparities between in-rack and off-rack bandwidth it could be a improvement to do something along these lines:
> {noformat}
> if (no. of known map-output locations > than no. of copier threads) {
>   try to schedule 75% of copies off-rack
>   try schedule 25% of copies in-rack
> }
> {noformat}
> This could lead to better utilization of both in-rack & switch b/w...
> Clearly we want to schedule more cross-switch than in-rack since off-rack copies will take significantly more time; hence the 75-25 split.

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