Alocação de infraestruturas virtuais utilizando algoritmos evolucionários multiobjetivo

• Ano de defesa: 2019
• Autor(a) principal: Souza, Wesley Oliveira
• Orientador(a): Salgueiro, Ricardo José Paiva de Britto
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Programa de Pós-Graduação: Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: Computação em nuvem; Alocação; Infraestruturas virtuais; Algoritmos evolucionários
• Palavras-chave em Inglês: Cloud computing; Allocation; Virtual infrastructures; Evolutionary algorithms
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://ri.ufs.br/jspui/handle/riufs/14547

To meet user requirements, Infrastructure Providers (InPs) began offering Virtual Infrastructure (VI) as a service. Among the tasks required to offer VIs as a service, the main one is the allocation of the requested VIs in the physical infrastructure. The allocation process consists of identifying within the infrastructure a feature set to host the components of the VIs. However, the allocation process is not trivial as it must meet predefined network and computing requirements. In addition, for effective infrastructure management, load balancing and reduction of allocation overhead is essential. Similarly, in the allocation process, some objectives of InPs and users should be considered. Generally guided by their financial perspective, InPs want to maximize their revenue by allocating as many VIs as possible using the smallest possible infrastructure. On the other hand, users mostly want efficient and low cost VIs. Therefore, the allocation process is complex and must meet a considerable set of constraints. To address this problem, the present dissertation presented an Evolutionary Multiobjective Algorithm (MOEA) to allocate VIs on a physical infrastructure, meeting computation and network requirements, evaluating solutions that meet goals such as load balancing and low allocation overhead. MOEA employed the proposed model for mapping the virtual resources of VIs into the physical resources of the physical infrastructure. In addition, a simulator was developed to evaluate solutions to the VI allocation problem. The experimental evaluation employed the simulation technique to evaluate the performance of the proposed solution. Thus, the algorithms were implemented in the Java language, and a comparative analysis was performed between different algorithms that employed the proposed allocation model. Thus, to evaluate the performance of the algorithms, the following metrics were used: time to fulfill an IV request, provider profit, rejection rate and efficiency of physical infrastructure use. In addition, a Cisco three-tier model-based topology was used to represent the physical infrastructure. The experimental results show that the developed Genetic Algorithm (GA) based MOEA presents promising results for several scenarios, combining speed and efficiency in the allocation. The proposed allocation model proved to be useful for mapping the virtual resources of VIs into the physical resources of the physical infrastructure. Therefore, the present work contributes to a possible solution to the problem and opens the way for new proposals that may employ the simulator and the proposed model.