Synthesis of a Reconfiguration Service for Mixed-Criticality Multi-Core Systems: An Experience Report

Task-level reconfiguration techniques in automotive applications aim to reallocate tasks to computation cores during failures to guarantee that the desired functionality is still delivered. We consider a class of mixed-criticality asymmetric multi-core systems inspired by our collaboration with General Motors, for which we automatically synthesize task-level reconfiguration services to reduce the number of processing cores and decrease cost without weakening fault-tolerance. We admit the following types of faults: safety violations by tasks, permanent core failures, and temporary core failures. We use timed games to synthesize the controllers. The services suspend and reinstate the periodic executions of the non-critical tasks to ensure enough processing capacity for the critical tasks by maintaining lookup tables, which keep track of processing capacity. We present a methodology to synthesize the services and use a case study to show that suitable abstractions can dramatically improve the scalability of timed games-based tools for solving industrial problems.