Performance debugging of heterogeneous real-time systems

Today most real-time embedded systems are made up of a heterogeneous collection of processors, communication subsystems and partially programmable or fixed-function components. These are typically Supplied by different vendors and as a result have different interfaces, require different programming models and implement different resource scheduling and arbitration policies. Hence, performance analysis and debugging Of Such systems is increasingly becoming complex. Although a lot of work exists in the real-time systems literature on timing and schedulability analysis of specific task and event models-which can be applied to analyze individual subsystems-the issue of compositionality has not received sufficient attention so far. In this paper we discuss a framework which can help in the analysis and performance debugging of such heterogeneous real-time systems. It can account for a variety of combinations of task and event models and scheduling policies and does not require any global state-space construction. As a result, it is highly scalable and can be used to analyze real-life hardware/software architectures. The main focus of this paper is on illustrating the utility of this framework in analyzing a heterogeneous collection of electronic control units that communicate via a FlexRay bus.