A Modified BCE Algorithm for Fault-Tolerance Scheduling of Periodic Tasks in Hard Real-Time Systems

Fault tolerance is an important aspect of real-time control systems, due to unavoidable timing constraints. In this paper, the timing problem of a set of concurrent periodic tasks is considered where each task has primary and alternate versions. In the literature, probability of fault in the alternate version of a task is assumed to be zero. Here, a fault probability with uniform distribution has been used. In addition, to cover the situations in which both versions are scheduled with some time overlapping, a criterion is defined for prioritizing primary version against the alternate version. A new scheduling algorithm is proposed based on the defined criterion. Simulation results show that an increase in the number of executed primary tasks which improves the efficiency of processor utilization, hence prove the efficiency of the proposed algorithm.