Record Length Requirement of LRD Traffic
Ming Li · 2022
This chapter discusses two problems in long-range-dependent (LRD) traffic measurement. One is how to determine record length of LRD traffic for a given estimation variance bound of a correlation or spectrum periodogram. The other is how to compute the estimation variance bound of a correlation or spectrum periodogram for a given record length and a given value of the Hurst parameter of LRD traffic. The solutions to the mentioned problems for general long-range-dependent stationary processes are given here. In addition, the solutions to the mentioned two problems specifically for the fractional Gaussian noise of the Weyl type and the generalized Cauchy process are presented. Moreover, real traffic in the Internet Archive by the Special Interest Group on Data Communication under the Association for Computing Machinery of US is analyzed in the case study. In this chapter, two types of periodograms, namely, the Schuster type and the Bartlett type, are considered. The results exhibit that either the periodogram of Schuster type or the Bartlett one can be used for correlation/spectrum measurement of LRD traffic.