Levenberg E. Evaluation of service quality characteristics at self-similar traffic of telecommunication networks

Object of study - processes of transmission and switching of packet traffic of multiservice communication networks. The subject of the study is methods of calculating the characteristics of service quality of self-service traffic of multiservice communication networks. Research methods. The research of the dissertation was made using the theory of queuing systems, teletraffic theory and probability theory. Using methods of approximation, mathematical statistics and consistency criterion, methods for calculating the self-similarity of packet network traffic are improved and a new probabilistic function of state distribution of a single-channel system with infinite queue is proposed; quality of service characteristics were evaluated using teletraffic theory. To evaluate the accuracy of the proposed solutions, simulation modeling methods were applied using mathematical statistics methods. Theoretical and practical results: uThe method of estimating the degree of self-similarity of packet traffic is improved by calculating the Hearst H index with the approximation of its nonlinear dependence on the parameter a = 1… 2 of the Pareto distribution for the first time; The method of estimating the degree of self-similarity of packet traffic by calculating the Hearst H index from the approximation of its first nonlinear dependence on the parameter a = 0… 1 of the Weibull distribution form has been improved; The method of calculating QoS characteristics of self-similar traffic in a single-channel system with infinite queue has been improved by using the Hearst indicator for this purpose; For the first time, an approximation of the probability distribution function of states of a single-channel system with infinite queue is proposed; The method of calculating the probability of waiting for service and the average delay of packets in the cumulative buffer of a single-channel system with infinite queue has been improved by applying for this approximation the system state distribution functions.