Aspect-Oriented Real-Time Architectures (Phase 2) (AORTA)

The goal of the AORTA project is to enhance the predictability of dynamic mixed-criticality real time systems by extracting critical paths. These paths are to be transformed into their static equivalents and to be executed in a time-triggered fashion at run-time. In comparison to event triggered processes, time-triggered execution tends to underuse resources. Therefore the optimistic execution model of mixed-criticality real-time systems will be retained. Only in case of an emergency the real-time system will be executed according to the static schedule. At the same time the results of the first funding phase will be generalized to dynamic real-time architectures. In particular, the focus will be on mixed-criticality systems with complex dependency patterns. The research project will investigate several variants of real-time Linux, as well as applications from the domain of control engineering.The main focus during the second funding phase of the project is going to be on dependencies between critical and non-critical paths of execution. These dependencies may potentially be problematic and can be found on all levels of the system: For example, application software may combine non-critical comfort functions with critical control functionality, leading to coupled components. In the operating system buffers may be used for shared communication stacks. Often such coupling may be desirable, however, in dynamic systems a host of possible execution paths at run-time may lead to dramatically overprovisioned system designs w.r.t. WCET and WCRT. Therefore, guaranteed execution times often lead to a loss of the efficiency gained from the dynamic real-time system design. Three key activities of this project will provide hard guarantees at run-time for the critical application core: analysis, tailoring and mechanisms.The basis for this project will be existing techniques for designing mixed-criticality systems under hard real-time constraints. For AORTA, it will be assumed that in general critical paths have deterministic structure and therefore their coupling with non-critical paths may be mapped to static equivalents. In the course of this project the applicability of the simple communication patterns provided by different variants of real-time Linux will be scrutinized to determine if these can guarantee the hard deadlines of safety-critical control applications, and if the concepts and techniques for static analysis, tailoring and scheduling developed in the first funding phase are suitable for this purpose. In addition the necessity of coupling the real-time architecture, scheduling and dependencies will be investigated in the context of mixed-criticality real-time systems to determine the general fitness of real-time Linux's design concepts for switching real time paradigms at run-time.