Openvisor – framework para redes de experimentação Openflow

• Ano de defesa: 2016
• Autor(a) principal: Powaczuk, Lucas
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; Brasil; Ciência da Computação; UFSM; Programa de Pós-Graduação em Informática; Centro de Tecnologia
• Palavras-chave em Português: Redes definidas por software; Redes de experimentação; Virtualização de redes; Software defined networking; Experimentation networks; Network virtualization; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.ufsm.br/handle/1/12942

OpenFlow-based testbeds have been established as an emerging field of research in order to create experimental environments that enable the development of new technologies on real network infrastructures. The bibliographic review showed that existing experimentation networks still lack mechanisms to guarantee users simplified operational forms, decoupled from the physical substrate and that are resilient. In this context, the research problem is: how to guarantee the users of OpenFlow experimentation networks an environment that allows creating virtual networks with low complexity in operation, flexible and resilient to link failures. The hypothesis that guided the study is that by integrating the tools OpenVirteX and FlowVisor and, consequently of its functionalities, the resulting framework would allow to achieving this purpose. OpenVirteX and FlowVisor are network hypervisors with distinct functionalities where the former has the use of virtual and arbitrary topologies, connectivity failure recovery, and absolute control. The FlowVisor has its main contribution in providing a wide flexibility in the definition of virtual networks. Therefore, the objective of this study was to develop a framework for OpenFlow experimentation networks, aiming to provide flexible virtual networks to users, with low complexity of the operation, having absolute control and resilient to failures. The study methodology is characterized by the hypothetical-deductive method. The procedures used to develop the proposal were: create the experimentation context, individual testing of the OpenVirteX and FlowVisor hypervisors, integration of the tools, evaluation of the framework and, finally, analysis and discussion of the results. The study confirmed some of the guiding hypothesis of the proposal since the framework was: Flexible, allowing to use any metrics of the OpenFlow header for the segmentation of virtual networks; Low complexity, because it allows to use a virtual and arbitrary topology composed of a single virtual switch corresponding to the entire physical network; Resilient to connectivity failures, because the tool was able to redefine the communication through of alternative routes. Regarding absolute control, the results refute the presence of this functionality. It was observed that providing total control of the network to the user has the impact of weakening the flexibility of the experimentation environment.