On the schedulability of real-time discrete-event systems

Abstract

We consider end-to-end latency specifications for hard real-time embedded systems. We introduce a discrete-event programming model generalizing such specifications, and address its schedulability problem for uniprocessor systems. This turns out to be rather idiosyncratic, involving complex, time-dependent release predicates and precedence constraints, quite unlike anything we have seen in the hard real-time computing literature. We prove the optimality of the earliest-deadline-first scheduling policy, and provide an algorithmic solution, reducing the schedulability problem to a reachability problem for timed automata.