Estimation of multifractal parameters in traffic measurement: An accuracy-based real-time approach

In this paper, we address the problem of real-time estimation of multifractal features for network traffic. The algorithm accuracy is the major concern in the proposed algorithm. From a statistical point of view, the higher the number of samples used in the estimation, the more accurate the results. However, network traffic in long intervals of time may have a heterogeneous scaling behavior, which would make the estimation results meaningless. We then propose an adaptive strategy that adjusts the length of the estimation interval based on local traffic features, i.e., it is enlarged as long as the traffic shows a homogeneous behavior. The development of this strategy relies on analyzing the variability of multifractality over time in real traffic traces. Simulation results show that the proposed algorithm is characterized by a higher accuracy with respect to a fixed approach.