Formal ADLs have significant potential for analyzing system architecture, predicting performance through simulations, and automating implementation parts. However, their industrial application is limited due to various challenges. This thesis addresses these challenges by designing a lightweight ADL that includes well-defined extension points for domain or process adaptation. It explores how architectural modeling can integrate agile development methods to facilitate incremental modeling and validation of system architectures. A comprehensive set of requirements for architectural modeling and simulation is established, leading to the development of MontiArc, an ADL for modeling logical architectures of distributed, interactive systems. MontiArc is grounded in the mathematical FOCUS framework, enabling event-based simulation of modeled systems. It includes code generators and a simulation framework for ongoing refinement and testing of architectural models. A method for language extension is introduced, allowing for the addition of new features and adaptation to different domains, enhancing the syntax, processing tools, and code generators. Additionally, a lightweight model library concept is proposed for the controlled and transparent development and reuse of component models and their implementations. The language, simulator, and extension techniques have been validated through multiple case studies involving MontiArc.
