Makespan and energy-aware workflow scheduler for heterogeneous cloud computing platform

Abstract

Scientific workflows, typically modelled as complex directed acyclic graphs (DAGs), are increasingly executed on heterogeneous cloud platforms to achieve high performance and scalability. However, as workflow sizes grow, energy consumption, and operational cost have become critical concerns, especially under global carbon-emission constraints. Although dynamic voltage and frequency scaling (DVFS) offers significant potential for energy savings, existing workflow scheduling methods fail to fully exploit heterogeneous processors that contain both high-performance and energy efficient cores, resulting in suboptimal makespan and energy utilization. To address this gap, the makespan and energy-aware workflow scheduler (MEAWS) is proposed as a multi-core DVFS-enabled scheduling framework designed to optimize both execution time and energy consumption in heterogeneous cloud environments. Extensive simulations using scientific workflows demonstrate that MEAWS reduces makespan by up to 88.75% and 70.4%, and lowers energy usage by 41.59% and 47.15% when compared with reliable and efficient workflow scheduling (REWS) and multi-objective workflow scheduling (MOWS). These improvements highlight the effectiveness of MEAWS in enhancing the sustainability and efficiency of scientific workflow execution.