CP-PGWO: multi-objective workflow scheduling for cloud computing using critical path

Abstract

When each task of the longest path in a task-dependent scientific workflow must meet a deadline, the path is called critical. Tasks in a critical path have priority over tasks in non-critical paths. Considering this fact that less methods have already dealt with the critical path problem for workflow scheduling in cloud, this study aims to present a critical-path based method to consider the problem based on our previous optimal workflow scheduling method, GWO-based (Grey Wolf Optimization). We applied our study to balance and imbalance scientific workflows. Our results show that considering the critical path improves the completion time of workflows while maintaining a proper level of resource cost and resource utilization. Moreover, to show the effectiveness of the current study, we compared the performance of the proposed method with non-critical-path aware algorithms, using three different indicators. The simulation demonstrates that compared to PGWO as the base method, the proposed approach achieves (1) approximately 68% improvement for makespan, (2) more accuracy in population sampling for about 70% of workflows, and (3) avoidance of the cost increases in more than 50% of workflows. Moreover, the proposed method decreases makespan approximately 3 times compared to the constrained-based approaches.