Enhancing Domain-Specific Software Architecture Recovery

Performing software architecture analysis and recovery on a large software system is expensive and time consuming; when it is done at all, it is often performed within a narrow context, focused on a few areas of particular concern. However, for a long-lived system within a well understood application domain, the costs for performing detailed architecture recovery may be amortized over several generations of the system; the resulting models can also be broadened and put into context by incorporating information about the history and anticipated future evolution of both the application and its underlying domain. This paper proposes a systematic approach for organizing application domain knowledge into a unified structure called the Architectural Domain Assets Set (ADAS). The ADAS structure builds on previous research, as well as our experience in performing an architecture recovery of IBM’s DB2. Our initial experiences in using ADAS suggest that it brings needed focus to the recovery process and provides assistance to domain-specific architecture recovery.