Database Replication applied to Network Management

• Autor(a) principal: Neves, Nuno Miguel Mendonça Correia das
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/61550
• topic: Software Defined Networking; Databases; Replication Mechanism; Network Monitoring; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Software Defined Networking (SDN) is a recent approach used to manage networks. Most of the time it is paired with OpenFlow, a low-level communication protocol used by controlling and switching devices to communicate. Since it is low-level, it does not grant the possibility to explore all the switching functionalities, especially as they get extended with more and more features. It is therefore required to find alternative ways of coordinating controlling and switching devices without resorting to low-level protocols to be able to access those functionalities. One of the possible approaches, which was recently implemented in a data center, uses databases and its respective replication to store and exchange information between the devices. Applying the same approach to manage wide area networks would provide a more flexible way to control them. The goal of this work consists of improving an existing prototype that simulates a small network. It was built originally using a SQL database and an asynchronous external replication software. We replace them with a NoSQL database that natively supports replication, which enables us to remove unnecessary software from the prototype while taking advantage of the database features. Some of the features, the more notable being non-uniform replication with the help of CRDTs, are used to improve network monitoring, which is a recent addition to the prototype. Network monitoring is a highly important component of network management that facilitates decision making processes. We evaluate the new version of the prototype by comparing with directly with the old version. We collect the convergence time of the network after an event on a device triggers a modification in its state to help with the comparison. By splinting the convergence time into a sum of smaller actions, we take conclusions regarding different moments of the convergence process.