ADORE-AR: Software Architecture Reconstruction with Partitioning and Clustering

The architecure documentation of a complex software system is a powerful enabler for the system.s successful development and evolution. Too often, the architecture has not been documented or the documentation has not been maintained. What should be done when an architecture description is acutely meeded? The answer proposed here is to reconstruct a description of the architecture that reconciles the expert.s understanding of the system and the current status of the system implementation. The dissertation focuses on how to reconstruct the architectural description of a system in an efficient manner, by combining tools and parsimonious expert input. It proposes a solution at two levels: ADORE-AR provides guidance for eliciting the architectural context and systems facts, reconstructing the appropriate architectural views, and then analyzing the resulting architecture, refining, and documenting it. SARA is a new semi-automatic subsystem decomposition technique, which interacts with the expert to converge quickly to a decomposition that he accepts. SARA uses partition algorithms to produce intermediate decompositions that optimize coupling and cohesion criteria