All of the previous discussion was spatial. This section will cover temporal decomposition. All formats should be able to use temporal decomposition as long as there is sufficient room on your servers so that each server can load the geometry and variables for one full timestep. Temporal decomposition is really just temporal caching of the timesteps before and after the current timestep. Each server reads one timestep and calculates all variables and creates all parts that you have created in your analysis on the current timestep. Therefore, changing timestep to the next (or previous) timestep occurs much more rapidly because the server that has responsibility that step has it pre-cached does not need to do I/O nor calculations nor part creation, and must simply pass its data up to the client. Note that the caching is for timesteps right around the current timestep: five servers and the current timestep is 6, might mean that one server is responsible for timestep 6 and the others are responsible for timesteps 4, 5, 7, and 8. Changing to timestep 12 will not benefit from the cached timesteps, and will, in fact, slow EnSight down: all five servers now will load five new timesteps simultaneously: 10, 11, 12, 13, and 14.

When loading multiple Case files into one EnSight Case, using the Multiple Case File (MCF) interface (see the Ansys EnSight User Manual: Reader Basics) and running in parallel with multiple Servers using temporal decomposition, there is no limitation on the number of Case Gold files relative to the number of servers because each server loads all of the data for its assigned timestep from all the files.