TUXUR - Um framework para divisão dinâmica de tarefas maleáveis em grade computacional

• Ano de defesa: 2014
• Autor(a) principal: Wiest, Roberto
• Orientador(a): Stein, Benhur de Oliveira
• Banca de defesa: Schnorr, Lucas Mello , Barcelos, Patrícia Pitthan de Araújo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria
• Programa de Pós-Graduação: Programa de Pós-Graduação em Informática
• Departamento: Ciência da Computação
• País: BR
• Palavras-chave em Português: Computação em grade; Framework; Tarefas maleáveis
• Palavras-chave em Inglês: Grid computing; Framework; Malleable tasks
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.ufsm.br/handle/1/5438

Grid computing, like other technologies of the Information Technology, was created and developed in the mid 90 in the academic scope. The idea of this technology is to use computer resources without worrying about acquisition new resources and physical location. The focus of computational grids is associated with the execution of applications that demand high computing power or adapt to distributed environments. This makes the environment dynamic and heterogeneous, characteristics that make resource management, scheduling and fault tolerance a great challenge. Grid applications are usually programmed through a framework of grid computing. In this work we present TUXUR, designed to manage the execution of tasks in computing grids. Theis framework is specialized in malleable tasks that allow being divided into independent subtasks of known size. The framework dynamically adjusts the size of the subtasks within the resources made available by grid. The main objective of this adjustment is to try to maximize the use of available computational power, balancing the workload of the components of the grid, according to the capacity of each one. The results of the evaluation show that this goal was achieved, maximizing resource utilization and decreasing the final execution time.