Analise e otimização de desempenho da Manhattan Street Network

AUTOR(ES)
DATA DE PUBLICAÇÃO

1996

RESUMO

We propose the use of buffers in the Manhattan Street Network (MSN) either to avoid or minimize packet losses and perform the packet loss analysis for the network with or without buffering. To have better understanding about the packet loss in MSN, we introduce a statistical model for the MSN with buffering. The model provides a probabilistic relationship between network link utilization and buffer occupation. ln the experimental studies, we compute some network statistics by solving the model as well as simulation. The results of the comparison among these statistics validates our approach. With the intention of optimizing the performance of The Manhattan Street Network, we propose a new routing algorithm. This algorithm is based on the Maxemchuk s routing algorithms and takes advantage of the use of queues to avoid the packet losses. We called Queueing Manhattan Street Network (QMSN) the MSN with queues and the new routing algorithm. ln order to evaluate the gain in performance obtained, we introduce a statistical model for QMSN, the One Node Model for QMSN, based on the One Node Model suggested by Greenberg e Goodman. Based on our model, we derive expressions for network performance statistics. ln experimental studies, we solved the model and obtained the performance statistics. The results, confirmed by simulations, prove that QMSN is a variant of MSN with better performance. A new concepts of network capacity and capacity rate are introduced to the Manhattan Street Network. An upper bound is found for the network capacity rate, which enables us to classify MSN into two categories: the normal operation region and the collpase region. The results from the packet loss analysis suggest that setting MSN to opearate below the network capacity rate guarantees the best trade-off among a high network throughput, low packet delay and light packet losses

ASSUNTO(S)

redes de computação redes locais de computação redes de comunicação redes de regições metropolitanas (computação) teoria das filas

ACESSO AO ARTIGO

Documentos Relacionados