Scheduling in a hard real-time system with shared resources

Abstract

Hard real-time systems are those systems in which correctness depends on logical results of the computations as well as the times at which these results are produced. In a real-time context, concurrency control protocols for shared resources (as databases) must not only maintain the consistency constraints but also satisfy timing requirements of transactions. So, several protocols have been proposed to prevent priority inversion. All these protocols, including the Priority Ceiling Protocol and the Stack-Based Resource Allocation Policy were initially designed to be used with the Rate Monotonic algorithm. We have implemented these protocols with their original scheduler, Rate Monotonic, and their dynamic version with the Earliest Deadline based scheduler, in order to compare their behaviour.<>