Skip navigation
Use este identificador para citar ou linkar para este item: https://repositorio.ufpe.br/handle/123456789/38329

Compartilhe esta página

Título: Utilizing optimization algorithms to maximize the availibility of composable data center
Autor(es): SILVA, Leylane Graziele Ferreira da
Palavras-chave: Redes de computadores; Algoritmos de otimização
Data do documento: 17-Fev-2020
Editor: Universidade Federal de Pernambuco
Citação: SILVA, Leylane Graziele Ferreira da. Utilizing optimization algorithms to maximize the availibility of composable data center. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020.
Abstract: The cloud computing paradigm has performed, for years, the fundamental role in delivering IT resources, typically available in data centers, allowing cost reduction and providing services such as high availability, scalability, and elasticity. Despite many advantages, the data center infrastructure suffers from inefficiency problems due to factors such as excessive redundancy usage and infrastructure sub-utilization. The Composable Data Center paradigm aims to mitigate such problems, proposing the disaggregation of resources distributed in racks with different chassis configurations. In this context, different resource arrangements, allocated via software (called Composable Infrastructure) may directly affect the system availability. Thus, this work presents an optimization problem to allocate Composable Infrastructures in Composable Data Center, taking into account budget constraints, obeying application requirements to maximizing availability in those data centers. For such, some optimization algorithms utilized in two approaches: mono-objective approach and multi-objective approach. From the results, it is possible to identify the best configurations and understand how each component affects availability. In the mono-objective approach, the Dynamic Programming algorithm obtained the best results in balancing cost and availability. In the multi-objective approach, both GDE3 and NSGA-II algorithms were useful in finding attractive solutions, and GDE3 presented the most solution number in most of the cases.
URI: https://repositorio.ufpe.br/handle/123456789/38329
Aparece nas coleções:Dissertações de Mestrado - Ciência da Computação

Arquivos associados a este item:
Arquivo Descrição TamanhoFormato 
DISSERTAÇÃO Leylane Graziele Ferreira da Silva.pdf1,85 MBAdobe PDFThumbnail
Visualizar/Abrir


Este arquivo é protegido por direitos autorais



Este item está licenciada sob uma Licença Creative Commons Creative Commons