An Aggressive Non-Preemptive Real-Time Scheduling Using Imprecise Computing
Xiangyu Wu, Kangda Zhang, Jerry Jerry · 2023
Many conventional real-time scheduling algorithms adopt pessimistic strategies to meet the deadlines in the worst case, which result in a waste of resource and energy. In order to solve this problem, we proposed an aggressive non-preemptive scheduling by MILP off-line formulation. By means of implementing imprecise computing on the algorithm and task granularity, satisfy the real-time constraints of aperiodic task set with a QoS guarantee. Furthermore, we proposed two preprocessing algorithms and three on-line adjustment schemes to improve different QoS factors at runtime. Simulation results show that these techniques can improve the schedulability of real-time systems with a high QoS. Considering the performance constraints of MILP off-line scheduling algorithm, it does not suitable for application in the large-scale task sets, a new heuristic algorithm to solve the problem is needed to explore. Besides, the execution time probability model should also be improved to enhance schedulability in the future.