cheduling scientific workflows of communicating tasks is the process of mapping tasks of the workflow to the cloud resources and directly or indirectly determining when to execute each task. Different objectives such as minimizing the makespan or price to be paid to the cloud provider, and respecting privacy are considered by different schedulers. Heterogeneous Earliest Finish Time (HEFT) was first designed to schedule communicating dependent set of tasks represented as a directed acyclic graph, on heterogeneous distributed systems. Because of its good performance in reducing makespan of running workflows, it is now widely used on the Grid and especially on the cloud. Based on HEFT, in this study, a static scheduling algorithm is designed to schedule scientific workflows on the cloud with the objective of improving HEFT’s performance with respect to the makespan criterion. With HEFT, when a task is ready to be scheduled, the resource that can complete it the earliest is selected. With the proposed method, when a task is ready for execution we assume that it is assigned to all leased resources to be executed. Then k more tasks from the priority queue of ready tasks are hypothetically scheduled with HEFT and whichever resource finishes the job the earliest is selected as the resource to run the task which we started with. The time complexity of the developed algorithm ant its impact on the reduction of makespan of number of real workflows are reported.

Keywords