Date of Award

Fall 2004

Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)


Electrical and Computer Engineering

First Advisor

Nirwan Ansari

Second Advisor

Kevin W. Lu

Third Advisor

Edwin Hou

Fourth Advisor

Lev A. Zakrevski

Fifth Advisor

Roberto Rojas-Cessa


The Resilient Packet Ring (RPR) is a new metro technology; RPR shares SONET's ability in providing fast recovery from link and node failures as well as inherits the cost and simplicity of Ethernet. RPR, like SONET/SDH, is a ring based architecture consisting of two optical rotating rings (uni-directional). In RPR, packets are removed from the ring at the destination so that different segments of the ring can be used at the same time for different flows; as a result, the spatial reuse feature is achieved. Enabling the spatial reuse feature introduces the challenge of guaranteeing fairness among the nodes sharing the same link.

The RPR fairness algorithm is comparatively simple, but it poses some critical limitations. One of the major problems is that the amount of bandwidth allocated by the algorithm oscillates severely under unbalanced traffic scenarios. These oscillations are a barrier to achieving spatial reuse and high bandwidth utilization. Moreover, the current RPR standard uses a single FIFO for each class at the ingress point, thus resulting in the head of line blocking problem. On the other hand, RPR uses the shortest path to route the traffic in the dual ring which is inefficient and unfair.

In this dissertation, the performance of the existing fairness algorithms and their limitations was investigated. Two bandwidth allocation algorithms were proposed to address the fairness issue. Both algorithms were demonstrated analytically and through simulations were able to achieve fairness and maximize the ring utilization. The Distributed Bandwidth Allocation (DBA) and the Adaptive Bandwidth Allocation (ABA) do not need to maintain information about each node. Instead, they use the local information which makes them scalable for a ring with any number of nodes. The Simple Scheduling Algorithm (SSA) was proposed to avoid the head of line blocking and to maximize the ring utilization at a very low complexity. The SSA algorithm was shown analytically and through simulations to be optimal where the flows achieve their max-mm fair rates at a very low computational complexity. Also, the weighted routing algorithm was proposed to maximize the ring utilization by enabling the RPR nodes to transmit in both rings in a weighted manner. The routing algorithm was demonstrated analytically and through simulations was able to maximize the ring utilization.