Resumen
The subject of the research is to predict the queue length for the communication device of a high-speed computer network with non-Gaussian traffic. The goal of this article is to examine the probabilities of the application of time scales used to study the organization of queues of modern high-speed computer networks. The following methods were used: the fractal analysis, scaling methods, methods of approximation. The following results were achieved: the results of the time scale selection for constructing adequate models of modern traffic were presented. The use of such models, in particular, enables studying the dynamics of the queues of active network devices, which is important for planning and distributing the network load. The use of statistical characteristics of traffic on a small number of time scales enables expanding theoretical concepts for critical time scales, which makes this approach applicable to any traffic process including the long-term traffic. In addition, the issues of describing the behaviour of queue tails for modern high-speed computer networks are considered and the properties of the proposed model approximations are determined. The analysis of the independent Gaussian model of a wavelet domain and the multifractal wavelet model showed the advantage of the first one for the fractal traffic and a slight discrepancy in the results for traffic close to the Gaussian one. Conclusions. Various approaches to the selection of time scales used in the study of the organization of queues of modern high-speed data networks were studied. The effect of the necessary accuracy and computational power required for calculating the maximum approximation were analyzed and it was established that exponential time scales are optimal for the fractal traffic. The impact of the tails of distributions in different time scales on the process of queue organization was also shown. It was noted that in the context of non-Gaussian traffic scenarios, the correlation structure (both short-term and long-term ones) does not describe the queues behaviour adequately enough.