Long-range Dependent Self-similar Network Traffic: A Simulation Study to Compare Some New Estimators

Abstract- The intensity of long-range dependence (LRD) of the communications network traffic can be measured using the Hurst parameter. There are various estimators of Hurst parameter which differ in reliability of their results. Getting reliable estimator can help to improve traffic characterization, performance modelling, planning and engineering of the real networks. This paper deals with the comparison of a few Hurst parameter estimators in standardized simulation experiment generating synthetic data sequences that exhibit long-range dependent features corresponding to observed data. The two best known classes of stationary processes with slowly decaying correlations (i.e., having long-range dependence) have been studied: fractional Gaussian noise (fGN) and fractional autoregressive integrated moving- average (FARIMA). Earlier authors introduced the estimator called “Hurst exponent from the autocorrelation function ” (HEAF) and in this work it is compared with an estimator based on the FARIMA process (FARIMA-H). Approximately unbiased versions are constructed and compared by simulation. I