Algoritmos evolucionários para o fatiamento de redes virtuais considerando a multidimensionalidade dos tipos de fatias

• Ano de defesa: 2023
• Autor(a) principal: Sousa, Rayner Gomes
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Mapeamento de recursos; Fatiamento de rede; Virtualização; Meta-heurísticas; Paralelização; Resource mapping; Allocation; Network slicing; Meta-heuristics; Parallelization; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.ufc.br/handle/riufc/75023

The network slicing is a new service in which the system responds to modern applications’ complex and heterogeneous requests. It is a vital technology for a communication infrastructure to serve different businesses. A network slice comprises a collection of resources and services that meet an application’s requirements, whether a more specific or broader business case. Slicing means lower acquisition and operating costs for service providers. A slice owner rents a portion of the network and does not need knowledge of how resources are instantiated, managed, and maintained. Network slicing uses a class of algorithms known as VNE. VNE maps a set of virtual nodes and links to a set of real nodes and links. This process is reduced to problems with NP-Hard complexity, so finding an efficient and practical solution is challenging. The sophistication of the demands of the new 5G applications, such as IoT, bring more complexity to the mapping process. Faced with the difficulties of the slicing problem, this work reveals our contributions, which are briefly: (a) we present a literature review allowing us to understand the current scenario, the advances and gaps in the area; (b) we design and describe the essential elements to simulate the slicing service; (c) we designed mechanisms to increase the number of slices of the mMTC type by taking advantage of the periodicity of the transmissions; (d) we plan new techniques for adapting meta-heuristics to network slicing problems considering the 5G system infrastructure; (e) we adapt the ED meta-heuristic to AG; (f) we designed two ways of parallelizing the ED to reduce the service time; (g) we adapt the metaheuristic AAA to deal with slicing problems with high node dimensionality; Finally, our work managed to maximize the acceptance rate of the network slicing service when compared to the traditional approaches found in the literature and we reduced the slicing time through parallelization techniques and combination of the node and link mapping process in a single function.