Existing performance analysis techniques have limitations when used on distributed virtual reality systems, including a lack of support for the measurement of interaction latency. Results tend to be numerical in nature, limiting their usefulness when comparing models. This paper describes an approach to using simulation to generate analytic forms The execution of each process is simulated, keeping track of the execution time of each process as a symbolic expression. Cycle times and latencies can be found by measuring the appropriate time intervals during simulation. The application of analytical simulation is illustrated by applying it to common issues in distributed virtual reality systems: network performance, database distribution and measurement of interactive performance. Comparing the analysis results with an implementation of the model demonstrates the accuracy of the predictions.