The study of self-similarity of the traffic transmitted in the backbone Internet channel

Abstract

The results of the research of information flows statistical characteristics in the Internet backbone channel are discussed. A distinctive feature of the study is the separation of a network traffic passing through the backbone channel into three private data flows that are formed by the information exchange sessions (sender-receiver bidirectional connection) of corresponding, depending on the volume of transmitted data, classes. Using the author's technique for each of the selected flows random sequences are obtained. They contain values of number of packets and data volumes transferred by sessions of the same class during a certain time interval (aggregation window). Distributions of these sequences and their Hurst exponents are studied. The received results led to the reasonable conclusion that the studied random sequences are not self-similar stochastic processes.