As organizations strive to keep up with the proliferation of new information technologies, the need to create and maintain a cohesive, effective, and extensible enterprise computing architecture assumes greater importance. An important part of designing an enterprise computing architecture is the estimation of local and global server capacities—a task made more complicated by the flexibility in partitioning client/server applications between desktop workstation and local or remote servers. This paper provides a simulation-based approach to estimating server processing and storage capacities that is driven by individual user requirements, work patterns, geographic location, and workgroup structures. The model supports alternative partitioning and allocation strategies. An application to the architecture design process for a medium-sized manufacturing organization is presented. Results from its application indicate the overall credibility of the model and the overall approach. The adoption of its recommendations in their architectural redesign effort testifies to its utility.