Scheduling Parallel Work ow Applications with Energy-Aware on a Cloud Platform

Abstract

An inefficient energy consumption of computing resources in a large cloud datacenter is a very important issue since the energy cost is now a major part of the operating expense. In this paper, the challenge of scheduling a parallel application on a cloud platform to achieve both time and energy efficiency is addressed by two new proposed algorithms Enhancing Heterogonous Earliest Finish Time (EHEFT) and Enhancing Critical Path on a Processor (ECPOP). The objective of these two algorithms is to reduce the energy consumption while achieving the best execution makespan. The algorithms use a metric that identifies and turns off the inefficient processors to reduce energy consumption. Then, the application tasks are rescheduled on fewer processors to obtain better energy efficiency. The experimental results from the simulation using real-world application workload show that the proposed algorithms not only reduce the energy consumption, but also maintain an acceptable scheduling quality. Thus, these algorithms can be employed to substantially reduce the operating cost in a large cloud computing system.