Comparative analysis of modern trends in the field of traffic models of data transmission networks

Abstract

To date, in matters of processing and managing network traffic, there is no single approach applicable to a wide pool of practical and applied tasks that would allow solving traffic management issues. Published works in this area are aimed at solving highly specialized problems: when applying complex solutions, these problems require the introduction of many additional parameters that increase computational complexity or solve only narrowly focused problems. This article provides a comparative analysis of classical network traffic models and reveals the possibility of practical application of such models in real-life problems. Classical traffic models are considered in detail, namely the Poisson model, heavy-tail traffic models, models based on Markov chains, traffic models based on the fractal theory and models based on stochastic time series. A mathematical description of each traffic model is also presented. Based on the results of the comparative analysis, the applicability of mathematical models to real projects was assessed. Based on it, two main problems were identified: first, the lack of consideration of the previous results of network traffic processing; secondly, the narrowly focused applicability of each of the models, given the rigid binding to subject areas, which allows solving only a narrow range of problems. The following indicators were taken as the criteria for evaluating network traffic models: the ability to scale the analyzed traffic, the ability to consider previous traffic data, computational complexity and the absence of some random features that could affect the operation of the model. A detailed study of the problem of traffic scaling revealed the main patterns, dependencies, dimensions of the traffic packet by the time it was processed.