Study of on‐line measurement of traffic self‐similarity

The research focuses on the analysis of university e‐learning network traffic to work out and validate the methods that are most suitable for robust analysis and on‐line monitoring of self‐similarity. Time series of network traffic analyzed are formed by registering data packets in a node at different regimes of network traffic and different ways of sampling. The results obtained have been processed by Fractan, Selfis programmes and the modules library SSE (Self‐similarity Estimator) developed in the paper, which employs the robust analysis methods. The methods implemented in the SSE (Self‐similar Estimator) have been tested by computer simulation applying the Janicki and Weron (2000) algorithm for generating random standard stable values. The research results show that the regression method implemented by the software modules library SSE is most applicable to the network traffic analysis. The investigation of traffic in the Siauliai University e‐learning network has been shown that the network traffic is self‐similar with the Hurst coefficient that changes in the interval [0.53, 0.70], the correspondent stability index changes in the interval [1.43, 1.89], the skewness not observed because the estimated β = 0.