Enhancing analysability and time predictability in UML/MARTE component-based application models

This paper presents how to integrate formally based models relying on the Synchronous Dataflow Model-of- Computation (MoC) in a UML/MARTE component-based application model. This model corresponds to a flexible and expressive modelling methodology, which facilitates finding and building an application model with the intended semantics, but does not help to ensure a strict fulfilment of functional and extrafunctional requirements. However, this capability has become a need in the context of modelling mixed-criticality applications. This paper shows how the component-based UML/MARTE models captured in such a flexible methodology can integrate parts transformable into synchronous data-flows. This formalism facilitates the analysis of the functional and extra-functional properties of such parts. Specifically, the paper shows a set of modelling patterns which can be translated into SDF counterparts. Relying on these patterns, a framework has been implemented which enables a bi-directional interoperability between the UML/MARTE models abiding the patterns and the formally-based ForSyDe methodology.