This paper discusses CAVEvis and a related set of tools for the interactive visualization and exploration of large sets of time-varying scalar and vector fields using the CAVE virtual reality environment. Since visualization of large data sets can be very time-consuming in both computation and rendering time, the task is distributed over multiple machines, each of which is specialized for some aspect of the visualization process. All modules must run asynchronously to maintain the highest level of interactivity. A model of distributed visualization is introduced that addresses important issues related to the management of time-dependent data, module synchronization, and interactivity bottlenecks.