Simulation and analysis of different traffic models for ATM networks

In the past few years both conventional and unconventional traffic models have been proposed as accurate models for traffic transported by asynchronous transfer mode (ATM) networks. More especially, most of these models have attempted to capture the fractal stochastic processes of real-traffic such as the long-range dependence, slowly decaying variance, power law spectral density and the Hurst effect properties. We simulated and compared the performance of some proposed models for the ATM input traffic where the ATM network was approximated by a G/D/1 queueing network with a general distributed arrival process, constant service process and a single server. The network parameters of interest were the cell loss probability and cell delay. The simulation tests showed that these models exhibit different queueing behavior for the same network parameters, whereas inter-arrivals based on exponential distribution did not capture burstiness over many time scales. We also discussed the implications of these traffic models to high-speed networks.