Improving architectural quality properties through model transformations

In this paper, we present an approach for refining software architectures to improve their quality properties using model transformations. The proposed approach is based on pattern-based model refactoring, i.e. use of well-known architectural patterns for restructuring certain elements in the software architecture in order to achieve specific quality properties such as improved performance, reliability, security, and scalability. First, the conceptual architecture view is modeled using annotated UML diagrams. A number of candidate architectures are considered as a starting point. Second, analysis models are derived from the UML models. Simulation and analysis tools are then performed on the analysis models to assess the quality properties. The result of the analysis can be used to advise the architect either to select the particular architecture out of several candidates or to refine the models in order to improve the quality properties. The feasibility of the approach for improved reliability is demonstrated by a case study.