Efficient Prediction of Makespan Matrix Workflow Scheduling Algorithm for Heterogeneous Cloud Environments

Abstract

Leveraging a cloud computing environment for executing workflow applications offers high flexibility and strong scalability, thereby significantly improving resource utilization. Current scholarly discussions heavily focus on effectively reducing the scheduling length (makespan) of parallel task sets and improving the efficiency of large workflow applications in cloud computing environments. Effectively managing task dependencies and execution sequences plays a crucial role in designing efficient workflow scheduling algorithms. This study forwards a high-efficiency workflow scheduling algorithm based on predict makespan matrix (PMMS) for heterogeneous cloud computing environments. First, PMMS calculates the priority of each task based on the predict makespan (PM) matrix and obtains the task scheduling list. Second, the optimistic scheduling length (OSL) value of each task is calculated based on the PM matrix and the earliest finish time. Third, the best virtual machine is selected for each task according to the minimum OSL value. A large number of substantial experiments show that the scheduling length of workflow for PMMS, compared with state-of-the-art HEFT, PEFT, and PPTS algorithms, is reduced by 6.84%–15.17%, 5.47%–11.39%, and 4.74%–17.27%, respectively. This hinges on the premise of ensuring priority constraints and not increasing the time complexity.