Alocação de taxa de transmissão utilizando predição do tráfego de rede baseada no expoente de Lyapunov e no parâmetro de Hurst
Nenhuma Miniatura disponível
Data
2021-02-25
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
This project describes a standard approach to analyze and measure experimental and synthetic data, which display linear and non-linear characteristics, within the spectrum of Chaos Theory and its canonical procedures, evaluating the performance of several algorithms. The main goal of this work is to propose a method to predict traffic and perform a dynamic traffic rate allocation for the network servers based on the Lyapunov exponent and the Hurst parameter, considering the analysis of a range of traffic samples and synthetic data, in order to quantify them with the use of mathematical methods that reveal their intrinsic features. Some of the characteristic processes described are self-similarity, long-range dependency among samples and multiscale behavior. Thus, it is necessary to: reconstruct phase space and the attractor with the ideal delay $\tau$, while describing a couple of methods to compute it; to determine embedding and correlation dimensions ($m$ and $D_2$, respectively); to calculate the Lyapunov exponent $\lambda$ and Hurst parameter $H$; to perform principal component analysis ($PCA$); to predict the traffic within the longest predictable duration constrained by the inverse of the Lyapunov exponent, and finally perform dynamic transmission rate allocation for the network servers. The simulations performed confirmed the efficiency of the proposed approach, regarding the sequence adopted in this work and the classification of data, especially in cases of mixed behavior between randomness and determinism.
Descrição
Palavras-chave
Citação
ROSA, É. R. C. Alocação de taxa de transmissão utilizando predição do tráfego de rede baseada no expoente de Lyapunov e no parâmetro de Hurst. 2021. 145 f. Dissertação (Mestrado em Engenharia Elétrica e da Computação) - Universidade Federal de Goiás, Goiânia, 2021.