A bandwidth balancing (BWB) mechanism has been added to the distributed queueing algorithm of the distributed-queue dual bus (DQDB) network. BWB can provide the requested throughputs by lightly loaded stations and evenly distribute the remaining channel bandwidth among overloaded stations. However, its operation requires the wastage of channel slots. The authors introduce a new bandwidth balancing mechanism for DQDB. The operation of the new mechanism requires one additional bit in the access control field (ACF) of the slot but has the advantage of exhibiting a similar behavior with the current BWB mechanism of DQDB without wasting any channel slots. For this reason, it can converge faster to the steady state in which fair bandwidth allocation is achieved. The authors investigate the throughput and delay performance of the proposed mechanism under one traffic class and examine its capacity to support multiple priority classes of traffic. They also compare its performance with the corresponding performance of the BWB mechanism of DQDB.>

An approach to the delay analysis of various service disciplines in symmetric token passing networks is presented. It is shown that exact or approximate accurate expressions for the average packet delay of different service disciplines can be directly derived by the delay expression of the exhaustive service system. This can be accomplished by inflating, by an appropriate factor, the packet size of the discipline whose performance is analyzed and then using the inflated packet in the delay expression of the exhaustive service system. The method is applied to the analysis of the limited service and TTS disciplines, and its accuracy is verified by simulation results. The advantages of the analytic method presented are its generality, higher accuracy over previous analytic methods, and insensitivity to network latency and other system parameters.>

This paper presents performance results that indicate that packetized voice service can be provided on a token-passing ring without adversely affecting the performance of data traffic. This is accomplished by introducing a relatively mild priority structure: stations are limited to a single packet transmission per medium access, and voice packets are given access priority over data packets at the same station. In addition, voice traffic is allowed longer packet lengths than data traffic. Several versions of this basic scheme are considered: 1) the number of active stations is constrained so that voice packets are guaranteed access within one packetization period, 2) no guarantee on access time is provided and voice packets are discarded when the waiting time exceeds one packetization period, and 3) no guarantee on access time is provided and voice packets are buffered until they can be transmitted.