Complementing model-driven development for the detection of software architecture erosion

Abstract

Detecting software architecture erosion is an important task during the development and maintenance of software systems. Even in model-driven approaches in which consistency between artifacts can partially be established by construction and consistency issues have been intensively investigated, the intended architecture and its realization may diverge with negative effects on software quality. In this article, we describe an approach to flexible architecture erosion detection for model-driven development approaches. Consistency constraints expressed by architectural aspects called architectural rules are specified as formulas on a common ontology, and models are mapped to instances of that ontology. A knowledge representation and reasoning system is then utilized to check whether these architectural rules are satisfied for a given set of models. We describe three case studies in which this approach has been used to detect architecture erosion flexibly and argue that the negative effects of architecture erosion can be minimized effectively.