Deadline-Aware Scheduling for Scientific Workflows in IaaS Cloud

Abstract

Efficient workflow scheduling is critical for achieving high performance in IaaS cloud. Although various types of workflow scheduling problems have been widely studied in a distributed environment, there are few initiatives to modify the IaaS cloud. However, the existing scheduling strategies failed to meet the QoS constraints and the resources utilization of the servers. In this paper, we develop a dynamic deadline-aware workflow scheduling (DAWS) strategy in the IaaS cloud. The algorithm devises an efficient strategy to calculate the sub-deadline of the tasks and deploys the tasks to the best-fit VM instances on the server to minimize the total execution time of the workflow. The DAWS algorithm also finds an optimal schedule of the tasks to deploy them optimally in the servers. This may minimize the makespan of the workflow while meeting the deadline. We simulate and compare the DAWS algorithm with the current state-of-the-art algorithms over various scientific workflows using various performance metrics in terms of makespan, SLR, throughput, reliability, and resource utilization.