Estratégias para o planejamento de redes ópticas de transporte

Abstract

In the thesis we present new strategies for the design of Optical Transport Networks (OTN). Starting with the physical topology, an equipment portfolio and a set of traffic demands we are able to place the equipment, solve the Routing and Wavelength Assignment (RWA) problem and determine the optical channels taking into account the network survivability. We seek to create realistic approaches aiming to bridge the gap between academia and industry in this context. We have considered the viability of each solution taking as reference the equipment characteristics and limitations, but always following the ITU OTN standards. Our strategies are computationally efficient with very low cost but able to minimize APEX and maximize the network availability. We have shown how the network design can be segmented without compromising the solution generality, allowing directly approach each subproblem. We have modeled more accurately the relationship between the number of amplifiers and the quality of the optical signal so now it is possible to balance the total number of optical amplifiers amplifiers and the signal quality at each receiver. The optical channel feasibility is determined in advance, before solving the RWA problem. For this problem a new and more flexible integer program model has been created. Called Multiservice RWA (MRWA), it was planned to deal with feasibility and routing information provided by prior steps. This also allowed minimize the required number of regenerators on the project with a efficient strategy. KEPLAN, a new OTN planning tool has been developed in order to aggregate all strategies in one software. It allocates equipment and creates an initial solution to the RWA with path protection, minimizing the amount of shared nodes into disjoint paths of minimum length. This was be done by a generalized Suurballe algorithm which will be shown. Our studies has shown that this tool has a very low computation cost. While the MRWA model was tested by simulating a set of 29 large real OTNs, where the optimality was achieved without great demand of time. Thanks to modern optimization techniques adapted to MRWA, it has able to optimally solve in less than 15 minutes a 100 nodes OTN.