Evaluating Software Architecture: A Systematic Mapping Study on Design Metrics and Their Applications

Abstract

Architectural design is pivotal in software development, shaping system functionality and behaviour. This study consolidates experiences with architectural design metrics to enhance software development processes and products. Metrics, while not definitive of quality, highlight potential issues and areas for improvement. Essential properties of effective metrics include simplicity, empirical credibility, consistency, uniform dimensionality, language independence, and actionable feedback. A systematic mapping study analysed publications on software architectural design metrics, focusing on quality models and metamodels. The study identified vital metrics such as coupling, cohesion, complexity, and modularity and explored their applications in various contexts, including microservices and security. Results indicate a lack of substantial experience in applying these metrics. The metrics taxonomy includes structural integrity, adaptability, complexity, maintainability, and traceability, providing a comprehensive framework for evaluating software architecture. The study emphasises the need for further research to develop predictive models using architectural metrics, which could improve software quality by proactively addressing architectural issues. Future efforts should delve into data accumulation and investigate models for using these metrics for predictive purposes, ultimately enhancing software quality and development processes.