[ https://issues.apache.org/jira/browse/YARN-4360?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15115851#comment-15115851 ]
Arun Suresh commented on YARN-4360: ----------------------------------- Spent some time going over {{StageAllocatorGreedyRLE}}. I generally makes sense.. I might be missing something, but : # Instead of doing a {{planModificiations.addInterval(..)}} in the {{trackProgress()}} method and then subsequently, remove the {{tempAllocations}} if the requests is not satisfied, why not just perform the {{planModification.addInterval()}} right before we return a successful allocation, in line 179 > Improve GreedyReservationAgent to support "early" allocations, and > performance improvements > -------------------------------------------------------------------------------------------- > > Key: YARN-4360 > URL: https://issues.apache.org/jira/browse/YARN-4360 > Project: Hadoop YARN > Issue Type: Sub-task > Components: capacityscheduler, fairscheduler, resourcemanager > Affects Versions: 2.8.0 > Reporter: Carlo Curino > Assignee: Carlo Curino > Attachments: YARN-4360.2.patch, YARN-4360.3.patch, YARN-4360.5.patch, > YARN-4360.patch > > > The GreedyReservationAgent allocates "as late as possible". Per various > conversations, it seems useful to have a mirror behavior that allocates as > early as possible. Also in the process we leverage improvements from > YARN-4358, and implement an RLE-aware StageAllocatorGreedy(RLE), which > significantly speeds up allocation. -- This message was sent by Atlassian JIRA (v6.3.4#6332)