Suportando a adaptação de aplicações pervasivas pelo uso de funções utilidade

• Ano de defesa: 2007
• Autor(a) principal: Leal, Diego Asth Schuenck
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Programa de Pós-Graduação em Computação; Computação
• Palavras-chave em Português: Processamento distribuído; Computação pervasiva; Computação móvel; Arquitetura de computadores; Auto-adaptação; Dimensões de QoS; Adaptação em tempo de execução; CR-RIO; Monitoração; Requisitos não-funcionais; Seleção de recursos; Função utilidade; contexto; Auto-configuration; QoS dimensions; Run-time adaptation; Monitoring; Non-functional requirements; Resource selection; Utility function; Context; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO::COMPUTABILIDADE E MODELOS DE COMPUTACAO
• Link de acesso: https://app.uff.br/riuff/handle/1/17111

Applications operate in dynamic environments, due to the variation of the availability of the resources or due to the variation of the user's preferences. To keep user's satisfaction during the operation of such applications, administration costs have grown up. Is this scenario, pervasive applications are aware of such variations in the operation context and adapt themselves without the action of users or administrators. This work aims to list the questions that have to be solved to make adaptation possible, among them: application's architecture, resource discovery, monitoring, user's preferences elicitation and, mainly, best available resource selection. Trying to solve this last question, we propose - as our main contribution - to use a utility function, as a technique which takes into account user's preferences (to weigh conflicting QoS dimensions) and the availability of the resources needed to provide the proper quality. An extra feature is the ease of implementation and the low processing costs, what makes it possible to be used in small devices, like PDAs. In order to validate our proposal, we present some examples where we used a utility function to select the best resource. In the example where a client selects the best AP, we analyzed real traffic traces of a corporate wireless network and traces generated by a simulator parameterized by realistic probability distributions. We verified the quality and robustness of a solution based in a utility function and concluded that the main QoS dimension commercially available to select the best AP, the signal strength, should not guide this selection exclusively.