As service compositions grow larger and more complex, so does the challenge of configuring the underlying hardware infrastructure on which the component services are deployed. With more configuration options (virtualized systems, cloud-based systems,
), the challenge grows more difficult. Configuring service-oriented systems involves balancing a competing set of priorities and choosing trade-offs to achieve high-priority goals. We describe a simulation-based methodology for supporting administrators in making these decisions by providing them with relevant information obtained using inexpensive simulation-generated data. From our existing services-aware simulation framework, we generated millions of performance metrics for a given system in varying configurations. We describe how we structured our simulation experiments to answer specific questions such as optimal service distribution across multiple servers; we relate a general methodology for assisting administrators in balancing trade-offs; and we present results establishing benchmarks for the cost and performance improvements we can expect from run-time configuration adaptation for this application.