You are viewing a plain text version of this content. The canonical link for it is here.

Posted to yarn-issues@hadoop.apache.org by "Tao Jie (JIRA)" <ji...@apache.org> on 2017/03/09 04:54:38 UTC

[jira] [Commented] (YARN-6307) Refactor FairShareComparator#compare

    [ https://issues.apache.org/jira/browse/YARN-6307?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15902482#comment-15902482 ] 

Tao Jie commented on YARN-6307:
-------------------------------

Thank you [~yufeigu], FairShareComparator#compare is called very frequently in each container allocation process. It would improve the scheduler performance if we can simplify this method.
Furthermore, I don't think it is necessary to sort the queue hierarchy from the root to leafqueue in every node update. Can we do the sort in the update thread, then share the result for node update? It would reduce much redundant sort. Maybe we can improve this in another JIRA.

> Refactor FairShareComparator#compare
> ------------------------------------
>
>                 Key: YARN-6307
>                 URL: https://issues.apache.org/jira/browse/YARN-6307
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: fairscheduler
>            Reporter: Yufei Gu
>            Assignee: Yufei Gu
>
> The method did three things: check the min share ratio, check weight ratio, break tied by submit time and name. They are mixed with each other which is not easy to read and maintenance,  poor style. Additionally, there are potential performance issues, like no need to calculate weight ratio every time. 



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org