Manikandan R created YUNIKORN-3233:
--------------------------------------
Summary: Improve victim selection algorithm
Key: YUNIKORN-3233
URL: https://issues.apache.org/jira/browse/YUNIKORN-3233
Project: Apache YuniKorn
Issue Type: Sub-task
Components: core - scheduler
Reporter: Manikandan R
Assignee: Manikandan R
After collecting the victims, traversal starts to pick each victim and add to
the final list until we meet the preemptable resource requirements. Sum of the
resources used by victims collected so far might fall below or exceed above the
preemptable required resources. It is very unlikely that both are same and
equals. To avoid “exceed above” cases, there could be a chance that victims
waiting next in the list might be good match when compared to the ones
processed earlier.
How do we pick more appropriate victims to match the expected preemptable
resources on “best fit” (resource wise) basis precisely to avoid freeing up
resources unnecessarily?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
