Performance analysis of a flexible protocol achieving user fairness in high-speed dual-bus networks with destination release

Performance unfairness has been an open problem in the IEEE 802.6 (distributed queue dual bus, or DQDB) protocol. The unfairness can be amplified when erasure capability is introduced. This study is motivated by some efforts made by others in attempting to balance throughput in DQDB networks. The authors investigate two more fairness criteria-packet delay and packet blocking. The three fairness criteria are studied together with the p_i-persistent protocol, assuming erasure capability in the network. For each criterion, values of various network settings which determine the achievability of the criterion are calculated analytically, and the performance characteristics of the protocol are shown. The analytical results closely match the results from simulations of the protocol. The results demonstrate that the fairness criterion on packet delay results in the best performance. However, the fair-packet delay requirement is also more difficult to maintain since it is the most sensitive to network load changes