Optimum: a MARTE-based methodology for schedulability analysis at early design stages

The construction of a design model is a critical phase in RTS development as the choices made have a direct impact on timing aspects. In traditional model-based approaches, the design relies largely on the designer experience. Once the design model is constructed, a convenient schedulability test has to be found in order to ensure that the design allows the respect of the timing constraints. This late analysis does not guarantee the existence of a test for the given design and does not allow early detection of unfeasible designs. In order to enhance reliability of RTS, this paper proposes a two step methodology for schedulability-aware real-time software design models construction. Having a high-level functional model as entry, the first methodology step consists in the workload model specification. The workload model represents the system end-to-end workload of critical scenarios triggered by an external stimulus and subject to hard-real time constraints. The second step is the refinement of the workload model towards a Schedulability Analysis Model defining an appropriate and analyzable threading strategy model. The threading strategy defines the set of units of execution taken into account by the scheduler of the system and their scheduling parameters. Different threading strategies can be then evaluated from a schedulability point of view in order to guide the designer towards the choice of the most appropriate one