Optimizing task scheduling for hpc using software defined network

Abstract

Features of cloud computing in terms of reliability, scalability, and resource pooling have inveigled scientists to deploy high performance computing (HPC) applications on cloud. Nevertheless, HPC applications have major challenges on cloud that emasculate gained benefits. This study aims to analyze challenges that are faced by HPC applications on cloud and focus on how the performance of cloud can be increased through a platform. In this paper, different approaches are discussed for improving HPC task1 scheduling performance on cloud using software-defined networking. Moreover, a task scheduling architecture is proposed to manage task scheduling of HPC applications, which can be later on used as a service to improve the profit of service providers. Bandwidth and capacity of virtual machines are the main parameters that are considered.