Traffic measurements and characterizations on HiPPI link, 16 Mbps token-ring and VISTAnet

With the fast growing of new networking technologies and new networking applications, the need for research on traffic measurement, traffic characterization and modelling has grown tremendously. The purpose of this research has been to address new traffic measurement issues that are raised due to the rapid changes in the communication networking environment. Three different traffic measurement experiments have been performed. Their traffic characteristics and traffic models are discussed. Statistical analysis techniques have been used in analyzing the traffic characteristics of the measured interarrival times, the number of arrivals, and the packet lengths. The first experiment is file transfer measurements on 800Mbps HiPPI links in a super-computing environment. A trend analysis model has been used to analyze the interarrival time trend of the HiPPI traffic. Our analytical result shows that the interarrival time trend exists for the HiPPI traffic. The second experiment is the internetworking LAN traffic measurement from a 16Mbps backbone token-ring in a very large working wide area network called the ISER network. A Markov modulated Poisson process has modelled the distribution of the number of arrivals from one traffic concentrator node to the other traffic concentrator node satisfactorily. The third experiment is for the radiation dose image traffic that is generated by a Cray-YMP supercomputer and transmitted through VISTAnet. Three measurements have been performed, and the three different models, namely a Mixed Erlang distribution with two branches model, a Deterministic distribution, and an Erlang distribution mixed with a Uniform distribution, have been proposed to model the interarrival time distributions. The model fitting results are fairly close to the measured interarrival time distributions.