Alocação de taxa de transmissão utilizando predição do tráfego de rede baseada no expoente de Lyapunov e no parâmetro de Hurst

• Ano de defesa: 2021
• Autor(a) principal: Rosa, Évelynn Regina da Costa
• Orientador(a): Rocha, Flávio Geraldo Coelho
• Banca de defesa: Rocha, Flávio Geraldo Coelho, Vieira, Flávio Henrique Telles, Souza, Marcos Antonio de, Cardoso, Alisson Assis
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Engenharia Elétrica e da Computação (EMC)
• Departamento: Escola de Engenharia Elétrica, Mecânica e de Computação - EMC (RG)
• País: Brasil
• Palavras-chave em Português: Dimensão de correlação; Reconstrução do espaço de fase; Expoente de Lyapunov; Parâmetro de Hurst; Predição de tráfego; Alocação dinâmica de taxa
• Palavras-chave em Inglês: Correlation dimension; Phase space reconstruction; Lyapunov exponent; Hurst parameter; Traffic prediction; Dynamic esource allocation
• Área do conhecimento CNPq: ENGENHARIAS
• Link de acesso: http://repositorio.bc.ufg.br/tede/handle/tede/11274

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.