Tool support for systematic class identification in object-oriented software architectures

Abstract

Software architectures have received considerable attention in both research and practice for representing system stakeholder concerns, and many researchers have leveraged object oriented models and methods for software architecture representation and evaluation. While the benefits associated with object oriented approaches are closely aligned with desirable qualities for software architectures (e.g., reusability, extensibility, comprehensibility, performance), these benefits are only realized through rational decision making by the analyst when identifying object oriented classes from a requirements specification. Traditionally, the class identification process has been somewhat subjective, with different analysts often arriving at completely different class models from the same requirements specification. Since this subjectivity does not lend itself to automation, little tool support is available for identifying object oriented classes and their relationships. The paper describes a tool under development, Reference Architecture Representation Environment (RARE), designed to systematically guide the analyst through class identification by applying heuristics associated with quality attributes and evaluating the resulting architecture based on relevant static metrics. RARE helps address a number of challenges typically faced during the class identification process, including: (1) developing an architecture that reflects the quality attributes prioritized by the analyst, (2) managing inherent conflicts between selected attributes, and (3) capturing analyst expertise and rationale for use by others.