Bandwidth allocation and fairness in resilient packet rings

This thesis investigates different aspects of bandwidth allocation and fairness in Resilient Packet Rings (RPR). RPR supports spatial reuse which increases throughput of the ring, however, it can result in unfairness in bandwidth allocation throughout the ring. Therefore, a rate-based fairness algorithm is employed at each RPR node to maintain fairness among nodes while achieving high throughput. We start by developing a new fairness reference model for RPR which can be used to examine fairness of a set of rate allocations. We show that our proposed fairness reference model is a general form of an existing one and allows each node to implement its own policy in distributing its share of bandwidth among its local flows to achieve different objectives. We then study and evaluate the performance of different policies which can be employed by an RPR node to divide its bandwidth share among its local flows. Afterwards, we model the fair rate calculation process in the RPR standard bandwidth allocation algorithms. The developed models provide insights on dynamics of these algorithms. We also study and analyze the effect of different system parameters on the performance of the these algorithms through analytical results. Finally we propose a new bandwidth allocation algorithm for RPR called Virtual Queueing (VQ) which aims at stabilizing the virtual queue model of each RPR node. We show that the RPR standard bandwidth allocation algorithms suffer from severe performance limitations. We also study the performance of another RPR bandwidth allocation algorithm called DVSR and show that this algorithm also exhibits low performance characteristics in a dynamic traffic environment. We then show that VQ outperforms these algorithms with lower computational complexity.