Um Método Eficiente e Preciso para Decomposição Temporal de Fluxos em Pacotes

Jose Carlos BelloraRosa Maria Meri Leão

Traffic modeling is essential for developing new network mechanisms, capacity planning, and traffic engineering. The parameters of the models, in general, require the computation of first and second order statistics from a packet trace. Nevertheless, the generation of packet level statistics, such as sizes and timestamps of individual packets, requires a complex measurement infra-structure and does not scale well with link speed. Flow level measurements can be more easily obtained than packet level statistics since routers offer tools that give information about the flow of packets that traverse them. We propose a method to generate a packet trace from flow level measurements. The results show that the packet traces obtained from the method reproduce the first and second order statistics of the real traces very accurately as well as the queue length distribution. We parameterize two traffic models using the estimated packet traces and evaluate the queue behavior, as an example of application.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: