Modelos para análise de disponibilidade de arquitetura de um serviço de Vod Streaming na nuvem

• Ano de defesa: 2015
• Autor(a) principal: BEZERRA, Maria Clara dos Santos
• Orientador(a): MACIEL, Paulo Romero Martins
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Pernambuco
• Programa de Pós-Graduação: Programa de Pos Graduacao em Ciencia da Computacao
• País: Brasil
• Palavras-chave em Português: Avaliação de desempenho; Computação em nuvem; Disponibilidade
• Link de acesso: https://repositorio.ufpe.br/handle/123456789/16819

Cloud computing is a computational paradigm that has become widely adopted in recent years. This is due to certain characteristics, in particular its ready scalability and the fact that it is generally offered to the client on a pay-per-use basis, through a type of cloud computing called Infrastructure as a Service (IaaS). However, the failure of components within the cloud is a fairly common occurrence that directly impacts on the availability of the service. Cloud providers attempt to meet the challenge of ensuring high availability by employing certain methods, such as sensitivity analysis to identify availability bottlenecks, and the inclusion of fault tolerance mechanisms to achieve greater reliability. From this perspective the current work proposes an effective availability analysis of a Video-on–Demand (VoD) streaming service based on the Eucalyptus cloud platform. The research was divided into two case studies. In the first study a basic architecture with no redundancy mechanisms was analysed. The methodology involved first modelling the system through a combination of continuous time Markov chains (CTMC) and reliability block diagrams (RBD), then performing an availability analysis, thereby achieving a model validation that represented the system’s behavior. A sensitivity analysis technique was also applied to identify availability bottlenecks in this basic architecture. Subsequent studies analysed availability in increasingly redundant architectures through the inclusion of warm-standby mechanisms. Changes to the basic design were guided by the sensitivity indices, and at the conclusion of the work a comparison was made between the availability values obtained for each system. A further case study analysed the service availability by making variations to the mean time of activation of the redundant mechanism. Results confirmed that the deployment of redundancy in cloud infrastructures in the form of warm standby mechanisms does lead to gains in system availability. It was also established that varying the activation time of the most reliable component in the system had a greater impact on the overall availability of the VoD service.