K. V. Logvinovich - Student, Siberian State Aerospace University named after M.V. Reshetnev Technician, JSC - Information Satellite System - Reshetnev Company?. E-mail: krisslogvinovich@bk.ru
A. V. Kuzovnikov - Ph.D. (Eng.), Associate Professor, Department of Space Information Systems, Siberian State Aerospace University named after M.V. Reshetnev Chief of Data Relay and Telecommunication Systems Department, JSC - Information Satellite System - Reshetnev Company?. E-mail: ujub@list.ru
Yu. G. Vygonskij - Associate Professor, Department of Space Information Systems, Siberian State Aerospace University named after M.V. Reshetnev Deputy General Designer, JSC - Information Satellite System - Reshetnev Company?. E-mail: vyura1@mail.ru

The rapid development of communications systems, the increase in number of users lead to the necessity of analysis of traffic loading and distribution. These days the main instruments of analysis of network loading are the autoregression models, self-similarity methods and connectionist network analysis. The review of existing published works [3-5] shows that the fractal analysis of network traffic is a promising technique. The objectives of this work are overview of basic models of self-similar traffic distribution and analysis of its effectiveness. The Hurst exponent [1, 2], heavy-tailed distribution [1, 2] and traffic simulation techniques in communications systems [1, 2] are used as the main principles of the theory of self-similar traffic. The development of telecommunications systems leads to the traffic increase and complication of traffic distribution methods. All this requires new methods for traffic distribution forecast in communications systems. The performed analysis shows how many measurements of entries and service rate are need to estimate a network loading, average holding time, average time of standing in a queue. As a result of this analysis the traffic distribution model and the calculations on basis of this model are made. The simulation modelling language GPSS (General Purpose Simulation System) is used for the self-similar traffic modelling. The results of calculations prove that network loading increases with an increase of intensity of entries of messages. Also the forecast of traffic distribution indicates a poor accuracy of exponential distribution method in comparison with the self-similarity method. Forecasting errors of loading distribution in communications systems lead to long queues in data transfer and a slowdown of a whole network.