User-oriented many-objective cloud workflow scheduling based on an improved knee point driven evolutionary algorithm

Abstract

Cloud computing is able to deliver large amount of computing resources on demand, and it has become one of the most effective ways to implement large-scale computationally intensive applications. In a cloud computing environment, applications typically involve workflows. Therefore, optimized workflow scheduling can greatly improve the overall performance of cloud computing. However, existing studies on cloud workflow scheduling usually consider at most three objectives only and effective methods to solve scheduling problems with four or more objectives still lack. To address the above issue, a new cloud workflow scheduling model is formulated that simultaneously considers four objectives, namely, minimization of makespan, minimization of the average execution time of all workflow instances, maximization of reliability, and minimization of the cost of workflow execution. To solve this four-objective scheduling problem, an improved knee point driven evolutionary algorithm is proposed. Extensive experimental results demonstrate that the improved algorithm outperforms existing popular many-objective evolutionary algorithms in most experimental scenarios studied in this work, in particular when there is sufficiently large amount of computing resource supply and the time for scheduling is limited.