Título:	Arquitetura multiagente com modelos de raciocínio distintos para gerenciamento de recursos em múltiplos provedores de nuvem
Autor(es):	Mendes, Aldo Henrique Dias
Orientador(es):	Ralha, Célia Ghedini
Assunto:	Otimização combinatória Programação linear Meta-heurística Computação em nuvem Computação em nuvem - monitoramento
Data de publicação:	13-Ago-2024
Data de defesa:	2-Fev-2024
Referência:	MENDES, Aldo Henrique Dias. Arquitetura multiagente com modelos de raciocínio distintos para gerenciamento de recursos em múltiplos provedores de nuvem. 2024. 126 f., il. Tese (Doutorado em Informática) — Universidade de Brasília, Brasília, 2024.
Abstract:	Scientific and commercial applications in cloud environments require diverse resources, resulting in the need for efficient management. The complexity of distributed systems in the cloud with different characteristics, technologies, and varied costs makes optimized resource management challenging. Multi-agent technology can offer significant improvements for resource management, with intelligent agents autonomously deciding on virtual machine (VM) resources. In this work, we propose MAS-Cloud+, a multi-agent architecture for the prediction, provisioning, and optimized resource monitoring in the cloud. MAS-Cloud+ implements agents with three reasoning models: heuristic, optimized, and meta-heuristic. MAS-Cloud+ instantiates VMs considering the Service Level Agreement (SLA) on cloud platforms, prioritizing user needs regarding time, cost, and resource waste, and providing appropriate selection for evaluated workloads. MAS-Cloud+ was evaluated with a Bioinformatics application for DNA sequence comparison with different workload sizes, executed on the AWS EC2 platform. A comparative study with Big Data applications using the Apache Spark Benchmark was performed on the AWS EC2 and Google Cloud platforms. The results with the Bioinformatics application indicate that the optimization model performs the best, while the heuristic presents the best cost. By providing a choice among various reasoning models, the results reveal that MAS-Cloud+ allows a more economical selection of instances, reducing the average execution cost of workloads by approximately 58% for WordCount, Sort, and PageRank. Regarding execution time, WordCount and PageRank show a reduction of 58%, while Sort shows an increase of 532.30%. The results indicate that MAS-Cloud+ is a promising solution for efficient resource management in the cloud.
Unidade Acadêmica:	Instituto de Ciências Exatas (IE) Departamento de Ciência da Computação (IE CIC)
Programa de pós-graduação:	Programa de Pós-Graduação em Informática
Aparece nas coleções:	Teses, dissertações e produtos pós-doutorado

Os itens no repositório estão protegidos por copyright, com todos os direitos reservados, salvo quando é indicado o contrário.