RotR: Rotational redundant task mapping for fail-operational MPSoCs

As transient and permanent failures are rise shrinking process technology, MPSoC systems with fail-operational behavior have become important, especially for safety-critical applications. We therefore propose RotR, a rotational task mapping approach for an active-redundancy-based system to facilitate parallel execution of redundant tasks. RotR maps tasks such that no single failure affects more than one copy of a redundant task, and utilizes a multi-functional voter task that adapts its functionality based on the system´s redundancy state after each component failure. RotR mapping and the proposed voter task jointly enable fail-operational behavior by seamlessly transitioning from higher reliability (e.g., Triple Modular Redundancy) to lower reliability (e.g., Double Modular Redundancy) without requiring task remapping. Our results show that RotR improves a system´s fault-tolerant lifetime on average by 37% and 48% over standard DMR and TMR systems, respectively. Furthermore, it improves the overall lifetime by 29% compared to the baseline system having no redundancy.