Caracterização e modelagem do comportamento de usuários de mapas Web para reprodução de carga de trabalho e avaliação de desempenho de sistemas baseados em tiles

• Ano de defesa: 2015
• Autor(a) principal: Braga, Vinícius Gonçalves
• Orientador(a): Cardoso, Kleber Vieira
• Banca de defesa: Cardoso, Kleber Vieira, Almeida, Jussara Marques de, Rocha, Ricardo Couto Antunes da, Rodrigues, Vagner José do Sacramento
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Ciência da Computação (INF)
• Departamento: Instituto de Informática - INF (RG)
• País: Brasil
• Palavras-chave em Português: Avaliação de desempenho; Sistemas de informação geográfica; Sistemas de mapa Web; Web map tile service
• Palavras-chave em Inglês: Performance evaluation; Geographic information systems; Web mapping systems; Web map tile service
• Área do conhecimento CNPq: CIENCIAS SOCIAIS APLICADAS::CIENCIA DA INFORMACAO
• Link de acesso: http://repositorio.bc.ufg.br/tede/handle/tede/5216

Web mapping systems, or Web GIS, are important tools for geographic orientation and spatial data analysis. In recent years, the use of these systems has increased, as well as the challenge to ensure performance while the number of users and the data volume continue to grow. The performance evaluation is an important activity to investigate and mitigate systems performance issues. The workload is the starting point of the performance evaluation and is responsible for sending resquests to the system under evaluation. The reliability of the evaluation results is related to an appropriate load. Despite the importance of web map systems, the literature as presented little efforts to model the workload of these systems. In this dissertation, we present a methodology to collect and analyze data in order to create a model of Web GIS users’ behavior and to instantiate the model in a workload generator. We also propose a generic model, named MUSe-GM (Maps User Session Generative Model), and present a characterization of the users’ behavior using data of the access to a popular mapping application, collected by an extension developed for the Google Chrome browser. The characterization results were used to develop an instance of the behavior model and to implement a workload generator. The instance was evaluated by testing in a realWeb map system, using the workload generator, and through simulations. The results were compared with two other models from literature. The proposed model in this dissertation was significantly different in several aspects compared to the other, presenting a behavior closer to the real users’ behavior.