Novel Approach for Task Scheduling of HPC Applications Using SDN

Abstract

AbstractModernizations and progresses in cloud computing are some of the developing trends in distributed computing that fascinated researchers to deploy high-performance computing (HPC) applications on a cloud. HPC applications in multitenant clouds with resource virtualization not only provide resource elasticity, but also response time for applications. In order to lower the time that is needed for task implementation in the process of data migration, task scheduling algorithm personates a vital role. This study aims to propose a novel approach for task and resource scheduling for HPC applications using software-defined networking. Herein, a precedence-based dynamical scheduling approach is provided to setup and update routes for concurrent flow of data during data transmission, so as to optimize resource hiring cost and minimize incoming applications’ execution time. Experimental results revealed that the proposed method offers considerably better computational resource utilization than classical scheduling methods like first come FCFS, round-robin and optimal task scheduling algorithm.KeywordsHigh performance computingSoftware-defined networkScheduling algorithms