Data Traffic Modeling—A Signal Processing Perspective
Athina P. Petropulu, Xueshi Yang · 2003
CONTENTS 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 6.2 Preliminaries-Self-Similarity, Long-Range Dependence, and Impulsiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 6.3 The Self-Similarity and Impulsive Nature of Data Traffic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 6.4 Modeling Impulsive Self-Similar Data Traffic . . . . . . . . . . . . . . . . . . . . . . 207 6.5 Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 6.6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Before the appearance of the Internet, teletraffic engineering had evolved around telephone communications. Telephone calls had been observed to arrive in a Poisson fashion, and the calls’ holding times were exponentially distributed. As the bandwidth requirements for a telephone call was fixed, meeting certain quality of service (QoS) requirement for telephone communication was equivalent to guaranteeing a certain call-blocking probability, which in turn could be achieved by appropriate allocation of network resources. The resource allocation could be carried out based on the classical queuing theory that is applied to a Poisson arrival process with exponential holding or service times.