Towards Realistic Core-Failure-Resilient Scheduling and Analysis

On uniprocessors, a failure of the single core means unavoidable system failure. However, on multicores, when a core fails, it is conceivable that the computation could continue on remaining cores in a degraded system mode indefinitely, until orderly shutdown and servicing can take place. This would be very desirable for critical applications but, apart from hardware and software support, it would require (i) a scheduling approach designed for providing such resilience and (ii) accompanying schedulability analysis, that derives offline the guarantees about the system meeting its deadlines at run-time, even if one core fails.