Arquitetura para clusterização de recursos baseado em seu poder computacional utilizando algoritmo hierárquico e assinatura comportamental

• Ano de defesa: 2018
• Autor(a) principal: Senger, Wagner
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Ponta Grossa; Brasil; Programa de Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: Assinaturas digitais; Algorítmos computacionais; Sistemas de computação em grade; Digital signatures; Computer algorithms; Computational grids (Computer systems); CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; Ciência da Computação
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/3745

In a Grid, resources with low computational power, when directly competing with each other in a environment, could be undersused, favoring in time its scape. To reduce this effect, and allow a better usage of the environment resources, improving the grid potential, this study proposes a method of elements organization, who mades all of the most important resources caracteristics could be used to create the groups. To make it, the amount of each relevant resources component to the target application are compiled, composing his computational power. To allow an equal support demand capacity, the resources are arranged in groups, and their allocation is based on the computational power calculated. The separation of the groups is performed by a hierarchical algorithm based on the most distant element, due to the balance between strong and weak computational elements that it provides, thus generating groups with more similar characteristics. This organization allows that when there is a need to distribute a task, any group is able to execute it, becouse all have an equivalent computational power. Nevertheless, the groups are distinguished by the use that each of their resources has, which is not repeated perfectly in other groups. Each resource has a usage pattern, which over time can be refined till it could be predicted, when a group is generated it starts to display time windows that represent the best moments in which the resources can be used. These windows are represented by a behavior profile called Behavioral Signature. After determined the signature of each resources group, the signature of the entire group is also determined, which reflects the availability of its resources. As all groups have similar conditions to execute a demand, the signature allows their distinction, showing the most opportune moment to use each one, providing a metric to choose the request destination.