Specifying Fault-Tolerant Mixed-Criticality Scheduling

This paper extends the ideas behind rely-guarantee conditions to cope with the specification of real-time schedulers that support Fixed Priority (FP) scheduling. The sporadic task model is considered in which tasks with defined minimum periods and worst-case execution times are assigned priorities; the shorter the period the higher the priority. At run-time the task with the highest priority executes preemptively. A Mixed-Criticality approach that partitions work between safety-critical event-triggered (ET) tasks and mission-critical time-triggered (TT) tasks is tackled. Previous work has considered faults arising from jobs executing for longer than expected; this chapter addresses arrival faults—run-time faults may occur when jobs from the same ET task arrive earlier than expected; the required fault-tolerant response is to compensate by executing the TT tasks less frequently. The specification is split over two phases, a Planning phase that confirms that the assumptions concerning the resource needs of the application can be managed on the chosen hardware platform and a run-time Scheduling phase whose specification combines rely-guarantee conditions, state-invariants and predicates to require that progress is made at a sufficient pace to ensure that all deadlines are met in both normal and fault-tolerant modes of operation. The scheduling phase is executed only if the Planning phase succeeds.