An ATM traffic model based on empirical traffic measurements

The characterization of ATM traffic is very important in many aspects such as validating traffic models, designing congestion control algorithms, implementing communication protocols, and developing switch architectures. In this paper, traffic descriptors such as peak rate and mean rate are calculated based on data gathered from the High-Speed Networking and Performance Lab (HSNP) at Michigan State University. These descriptors are used to define parameters for an ATM LAN/WAN workload model. A procedure for developing this workload model is presented. Cumulative distributions of the measured data are compared to those of simulated data as a way to validate the ATM workload model. A statistical test shows that there is no significant difference between the simulated data and the measured data. This workload model can be used to generate different types of application traffic in an ATM LAN/WAN. Such a model is also useful for studying communication protocols, scheduling algorithms, and congestion control schemes for ATM networks.