Donald Bren School of Information and Computer Sciences

This paper investigates the effects of protocol-processing overhead on the performance of error recovery schemes in high-speed network environments. The investigated error recovery schemes are:

• an edge-to-edge error recovery scheme, where retransmissions of erred packets only take place between source and destination nodes, and

• a link-by-link error recovery scheme, where retransmissions only take place between adjacent switching nodes.

For retransmission of erred packets, we consider both Go-Back-N and Selective-Repeat procedures in the analysis.

The performance measures we obtain are the distribution of transfer delays and the loss probability of packets across a network. To obtain these measures, this paper develops a tandem queueing network model with feedbacks where each queue represents a protocol layer within a switching node, rather than a switching node as a whole.

Numerical results show that for a network with very-high-speed/low-error-rate channels, an edge-to-edge scheme gives the smaller packet transmission delay than a link-by-link scheme for both Go-back-N and Selective-Repeat retransmission procedures, while keeping the packet loss probability sufficiently small.