Doubling runtime estimations to improve performance of backfill algorithms in cloud metaschedular considering job dependencies

Abstract

Job scheduling is a very challenging issue in cloud computing. Traditional backfill algorithms such as Easy and conservative are extensively used as job scheduling algorithms. Backfill algorithms require the shorter job to come forward if sufficient resources for the execution of this job are available and run in parallel with the currently running jobs provided it does not delay the next queued jobs. This technique is highly dependent on runtime estimations of job execution. Moreover in real life scenario it has seen that submitted job's may or may not be independent to each other. In this paper we have proposed a technique that uses dynamic grouping method to consider job dependencies and doubling runtime estimation method in cloud metaschedular to improve performance of backfill algorithm. Results have shown that doubling runtime estimations can significantly improve performance of backfill scheduling algorithms provided that the runtime estimations are correct. © 2013 Springer.