With distributed simulation, existing simulations, even from different disciplines, can be reused or made to interoperate. The efficiency of this technique is however not firmly established and it depends on the ability to satisfy a number of requirements, especially concerning simulation performance. The performance depends to a large extent on the structure and scaling of architectural components of the simulation execution platform.In this paper, we present an approach to address this problem: the design of models to capture the main characteristics of distributed simulations, run-time infrastructures and network architectures, and the development of a tool to predict performances. This tool, which is a simulator of HLA simulations is written by using HLA, and is evaluated on three test applications.