A Systemic Methodology for Software Architecture Analysis and Design

Abstract

The paper outlines a new systemic methodology for analysis and design of software architecture that addresses some major limitations in the current state of the art. First and foremost, current approaches fail to recognize the contextual environment as a major driver for software architecture. The higher abstraction level of software architecture is matched poorly by the lower-level, detail-oriented paradigms that the current architectural design methods employ. Worse yet, the current prescriptive specification methods often fail to handle adequately architectural evolution in the face of inherent contextual uncertainty and to scale to the size of modern software systems. We introduce a new software architecture methodology that approaches the software architecture domain with systemic techniques and we model explicitly contextual factors in software architecture definition. We introduce the concept of patterns for the contextual environment, which serve alongside architectural patterns as the primary vocabulary for architecture description and analysis. Our analysis approach uses a probabilistic modeling and decision formalism to guide software architecture evolution. Finally, we define an evaluation framework to assess the "fitness for purpose" of a proposed software architecture.