A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering

Thesis (M.Eng.) - Asian Institute of Technology, 1989

A single ring architecture functioning under the Token Ring protocol (basic Token Ring) has good delay and throughput characteristics under low loads. However, at high loads, the delay increases exponentially and the throughput saturates. The objective of this work was to explore methods which could possibly improve the performance of the basic Token Ring by giving reduced delay and increased capacity. To improve the characteristics of the basic Token Ring, an architecture providing multiple physical channels was found to be the alternative. As a bridged ring architecture satisfied this criteria, it was selected for performance evaluation under the environment, when the performance of the basic Token Ring deteriorated. The allocation of stations to rings dictated the performance of a bridged ring architecture. Therefore it was necessary that the configuration of the bridged ring architecture in terms of the number of rings and the station allocation in each ring be done based on the traffic between stations. An algorithm was postulated for the allocation of stations to rings with the goal of achieving the least delay and highest throughput. The algorithm worked with the constraint of keeping high-traffic station pairs on the same ring. Both the basic as well as the bridged Token Ring were assumed to conform with the IEEE 802.5 standard. They were modelled to work under symmetric traffic, non-exhaustive service discipline, fixed message length and single message token generation policy with source removal. They were then simulated to evaluate the performance. The multiple ring architecture was found to give a better performance than the single ring, by offering significantly reduced delays and yielding increased throughput for the same load. It was also confirmed that station allocation was a prime factor in the performance of a bridged ring architecture; because, the architecture suggested by the algorithm gave a better performance than a random allocation of stations to rings.