The p/sub i/-persistent protocol is based on a probabilistic scheduling mechanism (see Mukherjee and Meditch, 1988). The authors further develop the protocol to make it easily implementable, by allowing it to be sensitive to changing load conditions. They study various properties of a simple algorithm which stations execute independently by using channel feedback information. This results in a fully distributed control mechanism that continuously adjusts the station probabilities p/sub i/ at their proper levels as governed by the offered traffic. An extensive simulation model has been developed to study properties of this control mechanism such as p/sub i/ settling time and accuracy, behavior under step changes in traffic load, effect of injection of additional packets, and effect of various parameters associated with the underlying algorithm. These experiments indicate that this algorithm is suitable for implementing the protocol.< >

In the above-titled paper by Takine et al. (see ibid., vol.COM-36, no.10, p.1119-27, Oct. 1988) an exact analysis of a nonsymmetric polling system with single message buffers was reported. The commenters provide an alternate method for analyzing the exact mean waiting times of the individual stations in the same system by extending an exact analysis for the two-station case. Some corrections to the numerical results provided in the paper are made.< >

Several packet-switched, multiaccess protocols based on round-robin scheduling have been proposed for fiber-optic networks, e.g. D-Net, Expressnet, and Fasnet. The authors study simple modifications to these round-robin protocols in order to improve the channel capacity for fixed packet length. A representative example modification of D-Net based on generating periodic locomotives is called scheduled D-Net. A further modification, which combines the beneficial properties of unmodified and scheduled D-Net, results in an even more efficient mixed D-Net protocol. The authors outline, approximately model, and analyze scheduled and mixed D-Net protocols, and demonstrate quantitatively how these protocols outperforms unmodified D-Net. For this comparison, the protocols' delay-throughput characteristics are sued as the performance metric. Simulation results are also included