posted on 2021-05-23, 18:44authored byHossein Ghandehari-Alavijeh
Resilient Packet Ring (RPR) is a new Data Link Layer ring protocol. In RPR, the ring is a shared medium for multiple nodes compete to get a portion of shared bandwidth. Fairness algorithm is responsible for allocating fair bandwidth among competing nodes. In our research, we address the stability problems of the current RPR Fairness and introduce a new solution. The present work is the first control theoretic approach to RPR Fairness and Congestion Control that rigorously models the dynamics of RPR Fairness algorithm by using control theory. The key idea is to involve the active nodes in the Fairness and Queue Congestion Control process which means developing a decentralized control system. In RPR, when the number of nodes or the distance between the RPR nodes is high, the delay plays an important role in the behavior of the fairness which may lead to oscillation, instability and packet loss. We propose the implementation of Smith predictor as a valuable technique to overcome the effects of this delay and achieve higher throughput. Our new theoretical insights allow us to design fairness and congestion control algorithms that achieve fair bandwidth allocation and high throughput with small buffer requirement even in presence of large delay and large number of active nodes in the ring.