Multi-round real-time divisible load scheduling for clusters

Abstract

Quality of Service (QoS) provisioning for divisible loads in cluster computing has attracted more attention recently. To enhance QoS and provide performance guarantees in cluster computing environments for divisible loads, in this paper, we integrate a Simplified Multi-Round (SMR) strategy into the design of real-time scheduling algorithms for divisible load applications. Four contributions are made in this paper. First, we present algorithm SMR and extend it to compute a closed form formula for minimum number of processors required to meet an application deadline. Second, we derived a closed form solution for execution time of the optimized SMR. Third, we formally prove that optimized SMR results in better completion time than the single round strategy. Finally, we integrate SMR with our algorithm framework and propose two sets of efficient algorithms. © 2008 Springer Berlin Heidelberg.