Intelligent energy-efficient scheduling with ant colony techniques for heterogeneous edge computing

Energy efficiency is a significant issue in heterogeneous edge computing systems for a large number of latency-sensitive applications. This article presents an efficient technique to minimize energy overhead of time-constrained applications modeled by DAGs in heterogeneous edge computing. The technique is divided into three stages. First, we design a new method to compute task priority and propose the ant-colony based energy-aware scheduling algorithm to get a preliminary scheduling result. Second, taking the slack time between tasks and their deadlines into consideration, we propose the downward proportionally reclaiming slack algorithm to further cut down energy overhead by the DVFS technique. Third, taking the slack time between tasks into consideration, we propose the upward and downward proportionally reclaiming slack algorithm to cut down energy overhead by the DVFS technique again. Simulated results indicate that the presented technique is highly efficient in reducing energy overhead compared with state-of-the-art techniques using benchmarks of distinct characteristics. • We design a new method of computing task priority. • We make good use of the slack time to reduce energy consumption by DVFS. • We propose the DUPRS algorithm to reduce energy consumption via DVFS. • Simulated results indicate that DUPRS is highly efficient in reducing energy overhead.