Statistical characteristics of DIS traffic

The Advanced Research Projects Agency (ARPA) has initiated the Synthetic Theater of War (STOW) program to expand the current distributed interactive simulation (DIS) capability to handle larger exercises. Characterizations of the unclassified network traffic from the Synthetic Theater of War-Europe (STOW-E) demonstration are assessed in the context of self-similarity. At small time scales, the LAN traffic oscillates between burstiness and smoothness, whereas the WAN traffic exhibits smoothness. However, both the LAN and the WAN traffic exhibit high burstiness on larger time scales. Therefore, the traffic shows characteristics that would be produced by Poisson-type and self-similar processes. Additionally, in comparing the STOW-E traffic with an asymptotically self-similar process, we find that it has an unusually high Hurst parameter (i.e., H/spl sim/1). A simple mathematical model for the traffic is an uncorrelated process whose parameters vary according to another asymptotically self similar process of high burstiness. Since a combination of self-similar processes results in an asymptotically self-similar process, we conjecture that future STOW traffic should behave as asymptotically modulated self-similar processes.