The layered architecture recovery as a quadratic assignment problem

The layered architecture recovery as a quadratic assignment problem

Belle, Alvine Boaye and El Boussaidi, Ghizlane and Desrosiers, Christian and Kpodjedo, Sègla and Mili, Hafedh

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2015

Abstract : Software architecture recovery is a bottom-up process that aims at building high-level views that support the understanding of existing software applications. Many approaches have been proposed to support architecture recovery using various techniques. However, very few approaches are driven by the architectural styles that were used to build the systems under analysis. In this paper, we address the problem of recovering layered views of existing software systems. We re-examine the layered style to extract a set of fundamental principles which encompass a set of constraints that a layered system must conform to at design time and during its evolution. These constraints are used to guide the recovery process of layered architectures. In particular, we translate the problem of recovering the layered architecture into a quadratic assignment problem (QAP) based on these constraints, and we solve the QAP using a heuristic search algorithm. In this paper, we introduce the QAP formulation of the layering recovery and we present and discuss the results of the experimentation with the approach on four open source software systems.