10. Distributed Priority Assignment in Real-Time Systems

Recent advances in in-system performance analysis allow to determine feasibility of a system configuration within the system itself. Such methods have been successfully used to perform admission control for updates in distributed real-time systems. Parameter synthesis, which is necessary to complement the admission control with self-configuration capabilities, lags behind because current approaches cannot be distributed properly or due to necessary design-time preprocessing steps. In this chapter we present a distributed algorithm to find feasible execution priorities in distributed static-priority-preemptively (SPP) scheduled real-time systems under consideration of end-to-end path latencies. The presented algorithm builds on top of an existing distributed feasibility test, which is derived from compositional performance analysis [1]. With an extensive set of pseudo-randomly generated testcases we demonstrate the applicability of the approach and show that the proposed algorithm can even compete with state-of-the-art design time tools at a fraction of the run time. Thus, despite its application to admission control, the approach is generally applicable to the problem of scheduling priority assignment.