Strategy Switching: Smart Fault-Tolerance for Weakly-Hard Resource-Constrained Real-Time Applications

The probability of data corruption as a result of single event upsets (SEUs) increases as transistor sizes decrease. Software-based fault-tolerance can help offer protection against SEUs on Commercial off The Shelf (COTS) hardware. However, such fault tolerance relies on replication, for which there may be insufficient resources in resource-constrained environments. Systems in the weakly-hard real-time domain can tolerate some faults as a product of their domain. Combining both the need for fault-tolerance and the intrinsic ability to tolerate faults, we propose a new approach for applying fault-tolerance named strategy switching. Strategy switching minimizes the effective unmitigated fault-rate by switching which tasks are to be run under a fault-tolerance scheme at runtime. Our method does not require bounding the number of faults for a given number of consecutive iterations.

We show how our method improves the steady-state fault rate by analytically computing the rate for our test set of generated DAGs and comparing this against a static application of fault-tolerance. Finally, we validate our method using UPPAAL.