This paper presents an approximate analysis of the maximum throughput of an HDLC-based data link protocol for a bridged network. To model an end-to-end data link connection, which is connected across multiple transmission lines and bridge processors, a specific server, denoted as a bottleneck server, is defined. This is identified as the server (i.e., transmission line or processor) which gives the largest service time (i.e., transmission time or processing time) among all the servers over the connection.
The interval of receipt of information ( I ) frames at the receiver to that of the departure of corresponding Iframes at the bottleneck server is approximated. This approximation enables an analytical throughput expression to be derived by observing the bottleneck server only.
Compared with simulation results, analytical results obtained are in fairly good agreement if I-frames are discarded at the bridges located between the bottleneck server and the receiver. However, they are relatively overestimated if I-frames are discarded at the bridges located between the transmitter and the bottleneck server. It is shown also that the optimum window size is obtained as the smallest integer not less than the minimum acknowledgement delay over the service time of the bottleneck server. 14