Exploring optimal solutions for planning optical networks with graph coloring techniques

• Autor(a) principal: Gomes, Inês Ferreira
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/24243
• topic: Graph coloring; Greedy; Programação linear inteira -- Integer linear programming; Rede ótica -- Optical network; Tabu Search; Wavelength assignment; Atribuição de comprimentos de onda; Coloração de grafos

Optical networks are crucial in today’s communications, so their planning is of paramount importance. Routing and Wavelength Assignment (WA) are essential planning functions to transport the data in these networks. This work aims to study different graph coloring techniques for WA in static optical networks. We analyze and compare different graph coloring algorithms such as the Greedy heuristic, an exact algorithm based on Integer Linear Programming (ILP) and the Tabu Search meta-heuristic, for WA in optical networks. The studies are performed considering several real network scenarios, different logical and path topologies, and the number of colors and computation time of each one of the studied algorithms is analysed. We have concluded that the exact algorithm based on ILP, although giving always the optimum number of colors, is only applicable to networks with less than 20 nodes, due to its higher computation time. It was also concluded that the Tabu Search algorithm gives always the same results as the exact algorithm, for the real networks studied, but in a much faster computation time. Finally, we have concluded that the common Greedy algorithm performs as well as the Tabu Search algorithm, for all the real networks studied, but in some scenarios, with random path matrices, the Tabu Search gives a lower number of colors than the Greedy algorithm. For example, the Tabu Search gives less 35 colors than the Greedy algorithm for a random path matrix with 1000 paths and a 50% probability that each path has one or more links being used by the other paths.