Measuring system and software architecture complexity

Conventional measures of software code complexity, specifically code length, Halstead difficulty, and McCabe cyclomatic complexity, are adapted to measure architecture complexity. These measures are applied to architecture models based on Unified Modeling Language (UML) class, sequence, and state diagrams by redefining measure factors in terms of UML elements. This approach is illustrated using architecture diagrams from a real-time communication control system. This technique enables architects to consistently quantify system complexity early in and throughout the engineering lifecycle. This is an advance over both current software engineering practices that measure complexity after implementation and commercially available modeling tools that measure software code, not architectures.