[
https://issues.apache.org/jira/browse/YUNIKORN-807?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17402739#comment-17402739
]
Sunil G commented on YUNIKORN-807:
----------------------------------
Thanks [~ccondit]
Overall, I think this approach looks good and will definitely optimize our node
scorer algorithm.
Awesome performance numbers as well. Let's target for the next release.
A couple of high levels thoughts:
1. DRF implementation is poor in YuniKorn today. This will work as a good base.
And I think we can improve that in a follow up jira where we need to find
dominant or highly utilized resources in a node in a better way
2. Listeners model/approach looks good. Do you see a possible hotspot in the
nodeUpdated event handling?
Thanks for the PR and I will review more in PR.
> Improve performance of node sorting
> -----------------------------------
>
> Key: YUNIKORN-807
> URL: https://issues.apache.org/jira/browse/YUNIKORN-807
> Project: Apache YuniKorn
> Issue Type: Improvement
> Components: core - scheduler
> Reporter: Craig Condit
> Assignee: Craig Condit
> Priority: Major
> Labels: pull-request-available
> Attachments: Node Sorting Performance Improvement.pdf
>
>
> YuniKorn currently sorts all nodes on demand whenever scheduling of a
> container occurs. This causes significant performance degradation as the
> number of nodes increases.
> If we replace the on-demand sorting with a B-Tree sorted proactively, we can
> improve performance considerably.
> This is a similar approach to YUNIKORN-21, but without the associated
> behavioral changes.
> I've attached a design document with the details of the approach and the
> performance improvement gained.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
