Trustworthy components––compositionality and prediction

Abstract

This article defines key requirements for an architecture-based approach to trustworthy components. We then provide a brief overview of our architecture definition language RADL with a focus on compositionality and extra-functional properties. RADL aims at very high-level specification and validation of hierarchical assemblies of distributed real-time components. Several ideas in RADL are oriented towards modern middleware technologies such as .NET and EJB and to software-engineering methods such as UML. RADL dynamic models are centered around contracts, state machines and Petri nets. These are associated to contact points and connectors for defining connection constraints in architectural specification. They define configuration and behavioral contracts when they are associated to components and architectural assemblies of components. RADL contracts permit static compatibility checks and automatic gate adaptation for true black-box reuse. Dynamic monitoring of deployed components complements this with execution-based mechanism enabling prediction of extra-functional properties during architectural design.