Avaliação de diferentes métodos de inferência da demanda por banda do algoritmo de DBA no desempenho de redes XG-PON
Sandro Marcelo Rossi, Mariela Mayumi F. Sasaki

DOI: 10.14209/sbrt.2012.142
Evento: XXX Simpósio Brasileiro de Telecomunicações (SBrT2012)
Keywords: XG-PON DBA traffic prediction neural network
Abstract
This article presents a DBA algorithm and the comparison among four methods of determining one of its inputs, the bandwidth demand. For the performance evaluation, we used a XG-PON implemented in the OMNet++ simulation framework. The four methods are: conventional Status Report (SR) and Traffic Monitoring (TM), and TM-based prediction using a simple linear regression and a time-delay neural network. All evaluated methods have shown adequate QoS for triple-play services.

Download