To investigate classical scheduling schemes with power management in IaaS cloud environment for HPC workloads

Abstract

This paper addresses the well-known virtual machine (VM) scheduling problem in infrastructure as a service (IaaS) cloud environments to reduce the operational cost of the system together satisfying quality of service (QoS) constraints. We provide the study to investigate classical scheduling schemes incorporated with power management techniques i.e., the dynamic voltage and frequency scaling (DVFS) in Private IaaS cloud environment for high performance computing (HPC) workloads. Specifically, one first served-based policy i.e., first come first serve (FCFS), two size-based policies i.e., smallest job first (SmJF) and shortest job first (ShJF), and one backfilling policy i.e., aggressive backfilling first (AgBF), are considered to investigate the environment. The highlighted the strengths and weaknesses of the VM scheduling schemes in this paper can help to choose the most appropriate scheme in the environment.