: Prediction of the quality attributes of software architectures requires technologies that enable the application of analytic theories to component models. However, available analytic techniques generally operate on formal models specified in notations that cannot flexibly and intuitively capture the architectures of large-scale distributed systems. The construction of model interpreters that transform architectural models into analysis models has proved to be a time-consuming and difficult task. This paper describes (1) how a reusable model interpreter framework can reduce the complexity involved in this task, and (2) how such a framework can be designed, implemented, utilized, and evaluated.1 Key Words: Software architecture, model-driven engineering, component-based systems Category: D.2.2, D.2.10, D.2.11