Controle de admissão de conexão para servidores de vídeo sob demanda utilizando a teoria assintótica de muitas fontes

• Ano de defesa: 2010
• Autor(a) principal: Stahlschmidt, Thiago
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial
• Palavras-chave em Português: Telemática; Sistemas de transmissão de dados; Telecomunicações - Tráfego; Televisão na Internet; Engenharia elétrica; Telematics; Data transmission systems; Telecommunication - Traffic; Internet television; Electric engineering
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/1050

This work proposes a admission control architecture for video-on-demand servers, which provide deterministic guarantees on the maximum delay of video traffic and increase the network resource utilization. The framework take advantage of the fact that stored video streams can be processed off-line to calculate the space and time parameters in order to estimate its effective bandwidth under the many sources asymptotic regime. The automatic generation of video data representation (packetized and encapsulated in network protocols) in textual format allows a full characterization of the underlying process so as to precisely calculate its bandwidth requirements. To bypass the traffic prediction problem, since the off-line estimation does not capture the exact dynamics of the video stream in the time domain, the link is divided into several virtual links, where only streams with same content are multiplexed (group multiplexing approach). The introduction of a borrowing mechanism to distribute the excess bandwidth to other groups which request service increases the resource utilization and QoS. The admission control is aided by a Database Management System, where a stored procedure decides whether to accept or reject a connection by simple compare the sum of effective bandwidths for the traffic streams with the available resources. The efficiency of the model is analyzed and compared with deterministic approachs.