Model-Based Formalization of Software Architecture Knowledge

Abstract

Applying architecture knowledge promotes quality, reduces risks, and is crucial to best meet stakeholders' expectations. Current architecture knowledge is vast and ever-growing, however, it is also heterogeneous, diverse, disperse, and expressed at different levels of abstraction and rigor. In practice, architecture design is bounded by the architect's skills, experience, and the subset of knowledge he masters, and it also requires a huge effort to adjust such knowledge to the development scenario. The resulting architecture is therefore not as good as it could be. Moreover, the architect's effort is not repeatable as it is implicitly embedded in the architecture descriptions. Although model-based approaches are being used to capture particular domains and methods, most approaches lack either generality or homogeneity, making it hard to integrate, adapt and apply such knowledge. In this work, we use megamodeling to provide an homogeneous means for capturing architecture knowledge, making it shareable and reusable. We formally define a mapping from key architecture concepts to modeling artifacts. Also, an architecture design scripting language is used to capture the fine-grained design actions, making architecture design repeatable.