A Hybrid Job Scheduling Approach on Cloud Computing Environments on the Usage of Heuristics and Metaheuristics Methods

Abstract

The Information Technology Industry has been revolutionized through Cloud Computing by offering dynamic computing services to users through its on-demand provisioning of scalable and virtualized resources over the internet on a pay-per-use measured basis. Performance improvements in task scheduling can have a great impact on the efficiency of cloud computing. This paper proposes a hybrid task scheduling approach which employs the metaheuristic optimization technique, genetic algorithm to produce a certain combination of scheduling heuristics for processing cloud workloads. This approach is developed to optimize the performance metrics namely makespan, average flow time, throughput, and average waiting time. The developed approach is evaluated on the CloudSimPlus simulation framework using large-scale benchmarks against other heuristics in terms of the stated performance metrics. The results indicate that the proposed hybrid approach consistently outperforms the baseline individual heuristics in terms of the stated metrics irrespective of the scale of the workload. It is also observed that the optimization potential tends to increase as the workload scale becomes heavier and optimizing flow time produces complementary effects on the other metrics.