Sustentabilidade na nuvem: um modelo de eficiência energética em ambientes de computação verde

• Ano de defesa: 2024
• Autor(a) principal: REIS, Thiago Nelson Faria dos
• Orientador(a): TEIXEIRA, Mário Antônio Meireles
• Banca de defesa: TEIXEIRA, Mário Antônio Meireles , SOARES NETO, Carlos de Salles , PAIVA, Anselmo Cardoso de , SILVA, Francisco Airton Pereira da , CASTELO BRANCO, Kalinka Regina Lucas Jaquie
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO DOUTORADO EM CIÊNCIA DA COMPUTAÇÃO
• Departamento: DEPARTAMENTO DE INFORMÁTICA/CCET
• País: Brasil
• Palavras-chave em Português: emissão de carbono; computação verde; computação em nuvem; eficiência energética; inteligência artificial.
• Palavras-chave em Inglês: carbon emission; green computing; cloud computing; energy efficiency; artificial intelligence.
• Área do conhecimento CNPq: Ciência da Computação
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/5704

This thesis explores the growing relevance of cloud computing in everyday life and the business context, emphasizing the importance of proactive approaches to mitigate its environmental impact. Green cloud computing is an initiative aimed at reducing energy consumption and CO2 emissions associated with cloud, without compromising its functionality and performance. The main focus of this work is to evaluate the effectiveness of resource scheduling algorithms in cloud computing data centers and to develop an innovative methodology for calculating energy efficiency scores and classifying energy performance. Using the CloudSim Plus simulation environment, four algorithms - Round Robin (RR), Dynamic Voltage and Frequency Scaling (DVFS), Particle Swarm Optimization (PSO), and Ant Colony System (ACS) - have been compared through 800 simulations. In addition to the simulations, the methodology involved analyzing the data through rigorous statistical techniques, including the use of the T-student table, and creating an energy performance index derived from the results obtained. The research also incorporated artificial intelligence, specifically neural network-based classifiers, to improve the classification of energy levels. The results indicated a significant reduction in energy consumption and CO2 emissions, approximately 55%, and an improvement in the cost-effectiveness of virtual machine allocation, around 28%. This study demonstrates that the adoption of innovative scheduling strategies and the implementation of a quantitative energy evaluation model can significantly optimize cloud computing efficiency. Additionally, the proposal of a new score calculation and the creation of an energy performance scale offer valuable tools for optimization and sustainability in data centers.