An efficient algorithm for nonpreemptive periodic task scheduling under energy constraints

In this paper, we address the energy-aware DVS scheduling problems for non-preemptive periodic task sets in real-time systems. Based on Greedy method, we propose a novel heuristic, Average Power Directed (APD), to maximize the system value under energy constraints, as well as keep the system functional during the whole mission period. Under the energy constrained situation given in this paper, experiments on large number of task sets demonstrate that compared with Greedy method, APD method increased the system value by average 35.1%. Even compared with Genetic Algorithm (GA), the APD method achieves improvement by 7.3%. When the system becomes tighter energy constrained, the energy efficiency of APD increases rapidly within large scope, which proves the energy awareness of this method. APD method has high performance and linear complexity, which facilitate it an efficient energy-aware task scheduling algorithm.