Guaranteeing Fault Tolerance in Real Time Systems under Error Bursts

Real-time systems in dynamic environments can encounter faults that can lead to incorrect computation resulting in failure of system or exhibit degraded system performance. Earlier studies focused on managing systems susceptible to faults especially due to error bursts caused by environment exposure and can be disastrous. It is important to deploy strategies to guarantee schedulability of these systems using temporal redundancy for such error bursts. This research employs a methodology for determining upper bound margin of a fault burst model with Rate Monotonic (RM) scheduling, provide error recovery strategy under error bursts, worst case response time (WCRT) and scheduling feasibility. Analysis with a given task set shows the fault tolerant schedulability and an algorithm has been developed for schedulability under various fault burst durations. Evaluation with performance traces and WCRT have been carried out with GUI in MATLAB. This work can be extended to multiprocessor systems, tolerating multiple error bursts and spatial redundancy for safety critical systems.