Low power scheduling for periodic real-time systems with Dynamic Voltage Scaling processor

Energy-efficient task scheduling for periodic real-time systems has been extensively explored in past decades. Dynamic Voltage Scaling (DVS) techniques, known as an attractive method to trade the performance for reduced energy consumption, have been adopted by many previous studies to slow down the system when the workload is low. Most such studies utilize static off-line schemes with the assumption of worst-case execution workload for each task. Dynamic schemes have been developed to reclaim the slacks left by the earlier completion of tasks than their worst-case estimations. However, these algorithms used Earliest-Deadline-First (EDF) scheduling to guarantee the real-time requirement and seemed to be inefficient in reclaiming the slacks produced by the low-priority tasks. This paper presents a novel power-aware scheduling algorithm to reclaim more slacks and save more energy. It is shown by the experiment that the proposed algorithm can lead to more energy savings of up to 13% compared to EDF policy.