Adaptive Control Of Self-similar Internet Traffic Flow

We should analyze the characteristic of the internet traffic flow and build the model before we control it. In the past time many researchers suppose the access of internet traffic is Poisson distribution because of the limit of traffic measure technology and equipment. The conventional Markov models are not applicable to network traffic modeling. Through the research of LAN and WAN traffic, Leland and Klivansky found the self-similar nature of network traffic, but traditional network models can not describe this nature well. Self-similar is very important to system performance such as Cell Loss Ratio, network delay and so on, so study of self-similar models is increasingly significant nowadays. FARIMA(p, d, q) process can exactly capture both LRD and SRD behavior of a network traffic. Compared with other models, it is more appropriate to apply to the real network traffic data.We use FARIMA model to describe the traffic flow and regard it as the uncontrolled input(equaled to noise).As the dimension of difference operator is too high and difficult to control ,we reduce the model and transfer the model to the polynomial of noise. Then we use minimal variance control strategies to eliminate the error caused by noise, and make the length of queue stabilize at the threshold.We do some analysis and simulation and proves the availability of the strategies and stability of systems.