Traffic modeling with universal multifractals

A new approach to modeling long-range dependent traffic is proposed, based on the multifractal formalism. It consists of generating a fractal support to capture the intermittency of a bursty source, and a universal multifractal discrete cascade to model the traffic variability during the active periods. Model parameters are estimated from empirical traces, such as Web browsing, using multifractal analysis. A traffic trace was generated using the proposed model and was fed into a G/D/1 queue. It was found that the multifractal model outperforms other models in tracking the behavior of the system driven by actual traffic traces.