Tree-Based Overlay Networks (TBONs) provide a scalable communication and data manipulation infrastructure for tools and applications. There is a great deal of flexibility in mapping TBON processes to a parallel system's resources, but some process mappings provide better performance than others. In this talk, we present early work on an approach for predicting the performance of TBON-based tools and applications, with an emphasis on the impact of TBON topology on performance. Our approach uses discrete event simulation, and initially targets Cray XT platforms such as the Jaguar system deployed in the Leadership Computing Facility at Oak Ridge National Laboratory.