A queueing model for optimal segmentation of a token-ring local area network

As nodes are added to a token-ring LAN and compete for the token, the system may experience intolerable congestion or even become unstable. Following Husselbaugh (1995), we use a simple M/GI/1 queueing model to analyse whether a LAN with symmetric traffic should be segmented into subnetworks connected by a bridge. Our model accurately represents traffic as a function of the number of nodes in each segment and explicitly considers the trade off between reduced congestion and the cost of a bridge. We compare our model to the multi-class polling models in the literature on token-ring LANS.