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 "Sunil G (JIRA)" <ji...@apache.org> on 2017/07/25 18:34:00 UTC

[jira] [Commented] (YARN-6869) Resources.lessThan() et al should not use ResourceCalculator.compare()

    [ https://issues.apache.org/jira/browse/YARN-6869?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16100526#comment-16100526 ] 

Sunil G commented on YARN-6869:
-------------------------------

Thanks [~templedf]
While doing preemption, we ran into this issue. But I was thinking this api is needed a compare kind of method to get to know if there is any dominance or not for any resource type. Indeed it can have impact on these apis. If its a vanilla checks like lessThan, greaterThan etc, we can have api like absoluteLessThan, absoluteGreaterThan. And if dominance is mandatory to handle in these apis, we can leave these as it is. I am not having much of context regarding how important to see dominant check is needed for these api, so please feel free me to correct me. 

> Resources.lessThan() et al should not use ResourceCalculator.compare()
> ----------------------------------------------------------------------
>
>                 Key: YARN-6869
>                 URL: https://issues.apache.org/jira/browse/YARN-6869
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: resourcemanager
>    Affects Versions: 3.0.0-alpha4
>            Reporter: Daniel Templeton
>
> If you look at what the callers of {{lessThan()}}, {{lessThanOrEqual()}}, {{greaterThan()}}, and {{greaterThanOrEqual()}} expect the calls to do, there's no reason why the comparison should be done in context of the total resources.  Skipping the division and doing a direct comparison should improve performance.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

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