Adaptability of entity partitioning based on clustering in complex virtual testing system

To meet different real-time requirements between various testing entities, a complex virtual testing system was constructed on heterogeneous networks environment. Using the same clustering strategy under different networks could not meet the requirement of virtual testing system. Based on the network real-time simulation theory and the entities interactive graph, the coarse granularity of entities was defined for close coupling networks, and the fine granularity of entities was defined for loose coupling networks. The nonlinear clustering was proven to be better than linear clustering for fine granularity of entities, and the linear clustering was proven to be better than nonlinear clustering for coarse granularity of entities. The test and application show that this theory is feasible for choosing among entity merging algorithms and evaluating partitioning plans.