Compositional Scheduling Analysis Using Standard Event Models: the SymTA/S approach
Kai R. Richter · Digitale Bibliothek Braunschweig (Verbundzentrale Göttingen (VZG)) · 2004
Embedded real-time systems must meet a variety of timing requirements, such as deadlines and limited load or bandwidth. These properties depend heavily on interactions between tasks and on the scheduling of tasks and communications. Unfortunately, the current practice of specialization and re-use results in increasingly heterogeneous systems, which specifically complicates the scheduling analysis problem. Todays best practice of timed simulation is increasingly unreliable, mainly because the corner cases are extremely difficult to find and debug. As an alternative, a variety of systematic and formal approaches to scheduling analysis have been proposed. Most of them, however, are either limited to sub-problems, or use unwieldy and complex models that distract designers in practice. This thesis presents a novel, structured analysis procedure that a) can cope with the increasing complexity and heterogeneity of embedded systems, b) provides the modularity and flexibility that the established, re-use driven system integration style requires, and c) facilitates system integration using a comprehensible analytical model. The approach uses intuitive and standardized event models to represent the interfaces between different components and their scheduling. The clear interface structure allows -for the first time- the modular composition of heterogeneous sub-system analysis techniques. This provides designers with the flexibility to use their preferred scheduling and analysis techniques locally without compromising global scheduling analysis. This new analysis procedure has been implemented in the SymTA/S tool. As it can be efficiently applied in practice, it provides a serious and promising complement to simulation.