Variations of Conservative backfilling to improve fairness

Written with Avinab Rajbhandary (Knox `13) and Vitus J. Leung.
Proceedings of the 17th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2013.


Abstract:

We apply recent variations of Conservative backfilling in an effort to improve scheduler fairness. These variations modify the compression operation while preserving the key property that jobs never move later in the profile. We assess the variations using two measures of job-level fairness. Each of the variations turns out to be better than Conservative according to one of the metrics.