[
https://issues.apache.org/jira/browse/YARN-736?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13685854#comment-13685854
]
Karthik Kambatla commented on YARN-736:
---------------------------------------
Thanks Sandy. Few comments:
# The name ShareComputer doesn't capture the purpose of the class. What do you
think of ComputeResourceShares or something along those lines?
# In FairScheduler#isStarvedForDesiredFairShare(), the .5 in Resources.min()
should be made configurable.
# Add comments to capture the logic in ShareComputer/ComputeResourceShares and
FS#isStarvedForDesiredFairShare
> Add a multi-resource fair sharing metric
> ----------------------------------------
>
> Key: YARN-736
> URL: https://issues.apache.org/jira/browse/YARN-736
> Project: Hadoop YARN
> Issue Type: Improvement
> Components: scheduler
> Affects Versions: 2.0.4-alpha
> Reporter: Sandy Ryza
> Assignee: Sandy Ryza
> Attachments: YARN-736-1.patch, YARN-736.patch
>
>
> Currently, at a regular interval, the fair scheduler computes a fair memory
> share for each queue and application inside it. This fair share is not used
> for scheduling decisions, but is displayed in the web UI, exposed as a
> metric, and used for preemption decisions.
> With DRF and multi-resource scheduling, assigning a memory share as the fair
> share metric to every queue no longer makes sense. It's not obvious what the
> replacement should be, but probably something like fractional fairness within
> a queue, or distance from an ideal cluster state.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
