A framework for specification and analysis of software architectures

Software architecture is the abstract design of a software system. It plays a key role as a bridge between requirements and implementation, and is a blueprint for development. The architecture represents a set of early design decisions that are crucial to a system. Mistakes in those decisions are very costly if they remain undetected until the system is implemented and deployed. This is where formal specification and analysis fits in. Formal specification makes sure that an architecture design is represented in a rigorous and unambiguous way. Furthermore, a formally specified model allows the use of different analysis techniques for verifying the correctness of those crucial design decisions. This dissertation presented a framework, called SAM, for formal specification and analysis of software architectures. In terms of specification, formalisms and mechanisms were identified and chosen to specify software architecture based on different analysis needs. Formalisms for specifying properties were also explored, especially in the case of non-functional properties. In terms of analysis, the dissertation explored both the verification of functional properties and the evaluation of non-functional properties of software architecture. For the verification of functional property, methodologies were presented on how to apply existing model checking techniques on a SAM model. For the evaluation of non-functional properties, the dissertation first showed how to incorporate stochastic information into a SAM model, and then explained how to translate the model to existing tools and conducts the analysis using those tools. To alleviate the analysis work, we also provided a tool to automatically translate a SAM model for model checking. All the techniques and methods described in the dissertation were illustrated by examples or case studies, which also served a purpose of advocating the use of formal methods in practice