Development and analysis of a hybrid random/token access protocol for high speed local area networks

Abstract

In this dissertation, we develop and analyze a local area network (LAN) protocol aimed at high data rates and that supports integrated traffic. We make use of a unidirectional dual bus as the network topology and develop a medium access control (MAC) layer that is a hybrid of random access and implicit token passing schemes. Due to its hybrid nature, the proposed MAC eliminates the single station throughput problem of token passing protocols, while it retains high utilization under heavy loading. We describe the proposed medium access control in detail using state transition diagrams and discuss implementation aspects of the MAC. We then investigate performance of the proposed system via both analytic evaluation and Monte Carlo simulation. The maximum channel utilization and an upper bound on the access are obtained under a heavy load assumption. An approximate analytic model of the proposed system under light to medium loading is developed, and the channel utilization and the average access delay are obtained using a Laplace transform method. Although the analytic model is based on several unrealistic assumptions, it appears to retain the basic characteristics of the proposed system. The performance characteristics of the proposed system are investigated and compared to those of Expressnet via simulations. Unlike other hybrid systems where performance rapidly degrades as load increases, the proposed system outperforms Expressnet over a wide range of loads. It is also shown that the proposed system provides more bandwidth to stations performing file transfers when compared to Expressnet. For instance, under the proposed system, five clustered file transfer stations achieve a channel utilization of 0.75 whereas they achieve only 0.53 under Expressnet. We consider three methods which improve the utilization-delay performance, namely, channel selection, packet removal at the regenerating station, and two-way token passing under heavy loading. The proposed system is extended to support integrated traffic more effectively. To do this, a priority scheme is incorporated into the proposed MAC with minor modifications. The maximum number of isosynchronous sources satisfying a delay bound and the minimum guaranteed bandwidth allowed to asynchronous sources are derived. The priority scheme of the proposed system is compared to those of Expresssnet and FDDI via simulations for the integration of voice and data traffic. It is shown that the proposed system has a larger voice capacity than Expressnet and it is more flexible in assigning bandwidth to data stations than FDDI. For instance, the voice capacity of Expressnet is only 38% of that of the proposed system when the maximum allowable delay for the speech sample is 10 ms.