Investigation into real-time performance analysis and allocation of system resources to meet throughout and response time requirements has resulted in a proposed methodology for structuring software for real-time distributed systems. A queueing network model is used to estimate real-time performance for alternative software partitions, and those partitions which can satisfy user performance requirements while requiring minimal resources are determined. The current work is limited to feed-forward, pipelined structures. Partition blocks are formed by merging user-specified modules, and the required processing power to meet user requirements is then determined. A two queue model, representing module service and communications, is analyzed using queueing network approximation. The results show that the optimal solution set can be determined using a branch and bound approach. An example of the application of the results to a real system is provided. The results have been implemented as a user-interactive software structuring tool.