In this talk it is demonstrated that cooperative ARQ protocols have the desirable property of increasing the saturation throughput of fixed multiplexed access radio networks, subject to non-uniform offered traffic. The throughput gain is found to be increasing as the source-to-destination channel quality deteriorates and the source-to-relay channel quality ameliorates. Asymptotically, such gain approaches that of the optimum solution,i.e.,the perfect statistical multiplexing.

Current local recovery solutions perform fast recovery from single-failure pattern by making use of local signaling for fault detection and pre-computed paths to detour failed network elements. Besides single network element failures, today’s networks are increasingly facing a variety of additional failure patterns with multiple network element failures. Multi-failure patterns are more disruptive and require more backup paths. Fault detection capability has to be enhanced in that fast recovery from multi-failure patterns requires fast detection of non-adjacent network element failures besides those adjacent failures. The objective of this research is to adapt current local recovery solutions to deal with multi-failure scenarios, while retaining as much as possible the simplicity and the fast detection feature. This objective is achieved by optimally grouping failure patterns into clusters. Fault detection, or fault localization into the grouped clusters, is achieved by establishing lightweight and fast hello message sessions along pre-computed paths. Multiple backup paths are computed in conjunction with these hello session paths. An appropriate set of backup paths are chosen depending on the detection results of the hello session paths.

The presentation will be divided into three parts: first a short presentation about UNICAMP, the OptiNet laboratory and our work. Next, I will show the work done during my Master's program on protection schemes using sub-graph routing in optical networks. Sub-Graph Routing Protection is a method that creates different failure-dependent protection paths for each connection. Some schemes were proposed and compared, aiming at eliminating unnecessary connection reallocations on a failure occurrence, without compromising the acceptance of new connections. Finally, I will explain the work done during my first year in the Ph.D. program. It is about bandwidth sharing among non-cooperative users in a circuit-switched network, each of which may have his own quality-of-service requirements, expressed by its blocking probability. Two methods were analyzed for blocking differentiation. We have found closed expressions for the blocking probabilities in a single-link scenario, when classes have the same holding time. In the case where classes have different holding times the acceptance region can be defined, given the blocking probability limits.

Plug and play optical (PPO) nodes can be used to ease the deployment of optical networks. PPO nodes must self-adjust to changes of the optical physical topology and fiber propagation characteristics, and provide wavelength routing functionalities to the client layer. In the paper, a protocol is presented for physical topology discovery at the PPO node. It advertises available optical resources and changing conditions of the optical physical layer. The protocol is based on the flooding principle --- adopted in the open shortest path first protocol --- applied to optical transparency islands. As discussed in the paper, the proposed protocol provides an appealing solution to the problem of topology discovery and update in PPO node networks.