Based On Petri Net Method, Network Traffic Analysis And Simulation Study

Computer network now plays an increasingly essential part in modern society. In terms of recent research, old traffic model, for instance, the one bases on Poisson distribution, cannot explain the burst traffic phenomenon which contribute to the self-similarity feature of network traffic. Therefore, in this thesis, network traffic with self-similarity essence, is modeled by engaging Petri Net theory in the following two aspects:(1) Modeling and theoretic analyzing: An IP layer net with infinite buffer (Model Ⅰ) and a Ethernet with limited buffer (Model Ⅱ) are modeled with Petri Net components. We also analyze them by invariability-method and researchability-method.(2) Simulation: Specific software for Petri Net model is programmed to validate the new model. Different network performances due to different parameter sets of ON/OFF model are compared first. Then, our two new models are simulated to compensate defects in theoretical analysis. Some significant metrics, such as average queue length in Model I's buffer and network utilization of Model H, are statistically approximated as well.