The primary concern of this paper is delay analysis of IEEE 802.4 token-bus, IEEE 802.5 token-ring, and Fiber Distributed Data Interface (FDDI) when all the network traffic is of the highest priority class. We model them in an unified manner as a cyclic-service system with exhaustive time-limited service policies. We have specifically pointed out two variants of the time-limited service policies-nonpreemptive and look-ahead. The former models the token-bus, and the latter models the token-ring and FDDI. The major result of this paper is an iterative algorithm to approximate the mean delays for the nonpreemptive policy. Numerical results show that the algorithm is reasonably accurate for both deterministic and exponential frame sizes. Last, we apply the algorithm to the performance analysis of a token-bus. In the appendix, we also outline a similar approach for the look-ahead policy.>

The authors consider the IEEE 802.4 token-bus, IEEE 802.5 token-ring, and Fiber Distributed Data Interface (FDDI) network standards. They perform the delay analyses for these token-passing network standards when all the network traffic is of the highest priority class. They model them in an unified manner as a cyclic-service system with exhaustive time-limited service policies. They have specifically considered two variants of the time-limited service policies-nonpreemptive and look-ahead. The former models the token-bus and the latter models the token-ring and FDDI. The major result is an algorithm to approximate the mean delays for the nonpreemptive policy. Numerical results show that the algorithm is reasonably accurate for deterministic and exponential frame sizes. The algorithm is applied to the performance analysis of a token-bus. A similar approach for the look-ahead policy is outlined.>

Motivated by token-passing protocols in LANs, a system of multiple queues, attended by a cyclic server, is considered. The server will switch from the i/sup th/ queue to the next when either K/sub i/ customers are served consecutively or the queue is empty, whichever comes first. Using the pseudoconservation laws, an approximation method is developed to obtain the mean waiting time for each queue. The performance of this approximation is shown to be more robust and accurate then previous heuristic approximations. Further, it is shown to be suitable for modeling token-passing LANs.>

A multiqueue system with a cyclic server and a class of limited service policies is discussed. In particular, exhaustive limited (EL), gated limited, and general decrementing service policies are investigated. The derivations of the expected amount of work left in the queue at the server departures for these three policies are given. These results are all in terms of unknown boundary probabilities. Corresponding server vacation models are used to estimate these unknown probabilities. Pseudoconservation laws for these policies are subsequently obtained. Numerical results obtained for the EL policy are noted to be very accurate compared with simulation results. Conservation laws with mixed service policies, and exact expressions of mean waiting times in symmetric systems are given.>