Discovering Software Architectures with Search-Based Merge of UML Model Variants

Software reuse is a way to reduce costs and improve quality. However, in industry, the reuse of existing software artifacts is commonly done by ad hoc strategies such as clone-and-own. Clone-and-own leads to a set of system variants developed independently, despite of having similar parts. The maintenance of these independent variants is a difficult task, because of duplication and spread functionalities. One problem faced by developers and engineers is the lack of a global view of such variants, providing a better understanding of the actual state of the systems. In this paper we present an approach to discover the architecture of system variants using a search-based technique. Our approach identifies differences between models and uses these differences to generate candidate architectures. The goal is to find a candidate architecture most similar to a set of UML model variants. Our contribution is threefold: (i) we proposed an approach to discover model-based software architecture, (ii) we deal with the merging of multiple UML model variants; and (iii) our approach applies a search-based technique considering state-based merging of models. We evaluate our approach with four case studies and the results show that it is able to find good candidate architectures even when different features are spread among model variants.