Prediction of Self-Similar Traffic and its Application in Network Bandwidth Allocation

In this paper, traffic prediction models based on chaos theory are studied and compared with FARIMA (fractional autoregressive integrated moving average) predictors by means of the adopted measurements of predictability. The traffic prediction results are applied in the bandwidth allocation of a mesh network, and the OPNET simulation platform is developed in order to compare their effects. The adopted predictability measurements are inadequate because although the chaotic predictor based on the Lyapunov exponent with worse values of the measurements can timely predict the burstiness of self- similar traffic, the FARIMA predictor forecasts the burstiness with a time-delay. The DAMA (dynamic assignment multiaccess) bandwidth allocation strategy combined with the chaotic predictor can provide better QoS performance.