| Summary: | Disaster-based failures, due to either natural, technological or human causes, became more frequent in time and wider in scope, degrading drastically the communication services supported by telecommunication networks. This is of utmost importance since communication services are an important part of our society critical infrastructure. This issue is even more critical in optical networks where a single optical fiber can carry a very large amount of service demands. It is important not only to quickly recover the failed network elements disrupted by a disaster (post-disaster problem) but also to evaluate and minimize, before the disaster occurs, its impact on services between nodes outside the disaster area (predisaster problem). This thesis focuses on the pre-disaster problem and aims to investigate how Elastic Optical Networks (EONs) can be used in a cost-effective manner so that the impact of disaster-based failures is minimized. This thesis considers the disasters due to malicious humans attacks against some network elements. First, the vulnerability of existing optical networks to these failures is assessed, mainly resorting to variants of the Critical Node Detection (CND) problem, an optimization problem that seeks the set of nodes whose simultaneous failure mostly disrupt the network services. An exact row generation algorithm is proposed for the CND problem in the context of transparent optical networks. Then, we provide telecommunication operators with tools to enhance the robustness of optical network topologies against disaster-based failures. On one hand, we investigate how to upgrade existing optical networks, through link addition, aiming to enhance their robustness to multiple node failures. In a heuristic approach, within a given fiber length budget, a multi-start greedy randomized algorithm and a greedy deterministic algorithm are presented. In an exact approach, methods to compute the Pareto frontier of the bi-objective optimization problem that considers both the robustness maximization and the minimization of the cost of upgrading the network are proposed. On the other hand, a minimum cost gateway nodes selection to third-party networks with maximum disaster resilience against multiple failures is provided. The Gateway Node Selection (GNS) problem is defined, proposing an exact methodology to obtain all Pareto-optimal solutions. Finally, we exploit the advantages of spectrally (provided by elastic optical networking) flexible optical network planning and operations. Considering static and dynamic traffic, with a mix of unicast and anycast service demands, Routing, Modulation and Spectrum Assignment (RMSA) algorithms resilient to multiple node failures are proposed. These RMSA algorithms are based on an introduced metric, called path disaster availability, that measures the probability of a routing path not being disrupted by a multiple node attack.
|
|---|