End-to-end path computation schemes for traffic engineering in next generation multi-domain networks

This thesis proposes an overall framework for the end-to-end traffic engineered path computation problem. As discussed below, the framework is subdivided into three separate aspects, all relying on G/MPLS forwarding technology, which enables a controlled routing and the reservation of resources along traffic engineered paths.;The second aspect is the adaptation of the above mentioned distributed scheme into an inter-layer scheme for the consideration of joint multi-layer/multi-domain scenarios. Such scheme is necessary because most inter-layer traffic engineered path computation problems are also part of an inter-domain setting. The proposed scheme is applied within a complete traffic engineering solution. Moreover, the use of traffic demand forecasts for traffic engineering is evaluated. The proposed ideas are analyzed first by a comparative study and then through simulations on real world networks. The results compare the proposed scheme and its variants to current inter-layer methods. The results show that the proposed scheme performs better in terms of overall utilization and path setup time. Moreover, the results concerning the use of traffic forecasts clearly show that the accuracy of these demand predictions is not a factor in their usefulness within the proposed traffic engineering scheme.;The third aspect is the proposal of a novel constraint based shortest path computation algorithm which, for the first time, considers the adaptation capability of GMPLS nodes. The solution is based on a mathematical program. The constraints taken into consideration are specific GMPLS technological and traffic engineering best practice constraints. Indeed, the nesting/un-nesting capability of GMPLS nodes is considered, and the solution not only prioritized them over costly signal conversion, but also assures their correct ordering along the computed path (i.e., solving the parenthesis problem). The results obtained by solving the mathematical program validate its correctness. Then, the algorithm is simulated on real world networks for a large set of demands. The results undoubtedly prove its worth compared to existing proposals, in particular to a graph transformation method. (Abstract shortened by UMI.);The first aspect treated in this thesis consists in the definition of a novel inter-domain scheme that allows for the computation of inter-domain traffic engineered paths in a distributed manner among different administrations. The scheme relies on calculation nodes (PCEs) that can cooperatively compute inter-domain paths. The proposed solution respects both scalability and confidentiality requirements of inter-domain scenarios. Moreover, it establishes a pre-reservation procedure that enhances the effectiveness of the scheme with superior path deployment success rates. This is necessary because the time to compute an inter-domain path is usually longer, allowing fluctuations in networks resources. Accordingly, if resources are pre-reserved at computation time, it prevents their vanishing at path deployment time, and thus avoids blockage. The proposed solution is studied analytically and through rigorous simulations. The results prove that the proposed scheme allows for the optimal computation of inter-domain paths, and that the pre-reservation mechanism is beneficial when compared to the method without this mechanism.