A fault-tolerant scheduling algorithm for distributed control system with possible hardware and software faults
Liu Huai, Huang Jianxin · 2016
Distributed control systems (DCSs) are one class of the most important distributed real-time systems which are widely applied in many fields. With the increase of the number of processors and the complexity, a DCS is subject to hardware and software failures. In this paper, we present a fault-tolerant scheduling algorithm which can tolerate both hardware faults and software faults. The backup copy of a task is executed only when its corresponding primary copy fails due to a fault. In order to decrease the number of preemptions and lower the runtime overhead of the algorithm, non-preemptive Earliest Deadline First (EDF) is applied to schedule primary copies which are assigned to the same processor, and the backward non-preemptive EDF is applied to schedule backup copies and to calculate notification times of tasks. Compared with BCE and DPA, simulation results show that NEFT can obtain a higher success rate in executing primary copies and lower the runtime overhead of the algorithm.