Uma metodologia para caracterizaÃÃo de trÃfego de vÃdio baseada nos momentos e primeira e segunda ordem

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

This thesis presents a new methodology for video traffic characterization. A set of states are defined by a classification algorithm based on the first and second order moments of the codified video sequences. Modulation processes and time series generation that possess statistics associated with the states, defined by the classification algorithm considered in this work, originate a resultant series with statistics equivalent to the original sequence. The proposal distinguishes itself from the rest for its flexibility and adaptation to the traffic characteristics to be modeled. The statistical classification adopted determines states with enough small variability, making possible the use of traditional models. The approach allows the use of both markovian and fractional differentiation models for the generation of the characteristic series and modulation. The inherent variability of the video sources can be absorbed in the modeling from an appropriate choice of the involved processes. Analysis done through the generation of series for representative individual and aggregated video sources had shown the validity of our methodology. Statistical equivalences had been ratified by means of the comparison of their histograms and QQ-plots. Measurements involving the mean absolute percentage error and the Akaike s information criterion demonstrated that modulated process origins better results. The statistical multiplexing effect in a computer network, in terms of bits loss rate, was evaluated through simulation both for traffic proceeding from individual sources and for an aggregate of sources. The obtained results showed that the proposed approach can be used as an efficient tool for video traffic characterization and in studies involving resource allocation in computer networks, elaboration of rules for buffer dimensioning at the network switches, predictions of QoS degradation caused by congestion, and congestion control mechanisms

ASSUNTO(S)

geraÃÃo de trÃfego markovian and fractional models vÃdeos computer networks redes e sistemas distribuÃdos trÃfego de redes ciencia da computacao video traffic traffic generation modelos fractais e markovianos

ACESSO AO ARTIGO

Documentos Relacionados