End-host multicasting in support of distributed real-time simulation systems

There has recently been a tremendous level of interest in the use of end-host multicast to provide group communication within a wide area network. End-host multicast shifts the burden of group-communication support from network routers to the host. End-host multicasting is required when IP multicasting is not available. The difficulty in using this technique is that it requires additional resource consumption when compared to native IP multicasting. This work presents several novel end-host multicast algorithms designed for peer-to-peer real-time simulation systems. Our techniques utilize application level admission control in order to more fairly balance the communication load in the physical network. We propose and evaluate an adaptive routing policy to find overflow paths between end-hosts and simulation agents placed within the system. We also present several heuristics for routing tree construction, including a simulated annealing approach. We have evaluated our methods through simulation, and the results show that our methods achieve performance similar to optimal IP multicasting.