Time series models for Internet data traffic

Abstract

A statistical analysis of Internet traffic measurements from a campus site is carried out to examine the influence of the constituent protocols and applications on the characteristics of the aggregate stream and on packet loss statistics. While TCP remains the dominant traffic protocol through all hours of the day, a mixture of both well-known (HTTP, FTP, NNTP and SMTP) and less known applications contribute significant portions to the TCP traffic mix. Statistical tests show that the aggregate TCP packet arrival process exhibits both nonstationary and nonlinear features. By filtering a subset of the applications found to exhibit nonstationary features from the aggregate process, a stationary traffic stream is derived. This filtered traffic process is modeled using nonlinear threshold autoregressive processes. The traffic model is shown to provide good agreement with the measurement trace in the packet loss statistics. The proposed parametric model allows the design of traffic shapers and provides a simple and accurate approach for simulating Internet data traffic patterns.