Critical ethernet based on OpenFlow

• Autor(a) principal: Martins, André Alexandre Nunes
• Data de Publicação: 2014
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/13972
• topic: Engenharia de computadores; Ethernet (Redes de computadores); Computação em nuvem; Computação de tolerância de falhas; Redundância

Nowadays, we put an immense value on Ethernet networks, especially for data center operations empowering cloud environments or huge network infrastructures in general. However, it is not always possible to bring 100% up-time communications since redundancy in Ethernet has always been an unresolved problem, considering the large amount of network resources to be managed. Through history there have been many developed solutions that tried to solve this issue, only to fail in providing the proper support. Software-defined Networking (SDN) is a novel paradigm and a dynamic and configurable mechanism that brings a programmable nature for developers to implement solutions that may finally solve the identified issues. Via the use of programmable open interfaces, the control and management of network behavior is becoming easier and less error prone. The main objective of this dissertation was the implementation and evaluation of a fail-safe SDN-based solution for critical communications, therefore for fault management in redundant Ethernet technologies on a typical data center management scenario. This dissertation presents the developed solution and the main phases of its implementations. The implemented solution uses a redundant L2 network and a SDN controller to calculate the network topology. The solution makes use of extensions to both the OpenFlow protocol and OpenDaylight controller’s modules. During the evaluation stage, different scenarios were tested where topology changes occur. The evaluation results show that the proposed solution behaves satisfactorily whenever a link fails, obtaining none packet loss. To conclude, the solution shows to be promising for critical data center operations concerning the adaptation time obtained.