Techniques for Failure Recovery in a Software-Defined Network

Abstract

As our lives become ever more dependent on network connectivity, it becomes increasingly more important for networks to be able to overcome the failure of individual components and continue to function. This thesis examines approaches to fault tolerance in software defined networks, and how the global viewpoint that Software-Defined Networking provides can be leveraged to create more reliable networks.

In order to continue operation after the failure of a network component, the failure must first be detected, and then the network must automatically change its behaviour to mitigate any adverse consequences. This thesis evaluates a variety of fault detection methods and potential responses. Based on these evaluations the design for a fault tolerance system for software defined networks is presented. This system builds protected paths using Ring Based Forwarding, an algorithm for creating a full mesh of paths between switches in a network where each path has a fail-over path at each hop. The system monitors the network for faults using Traffic Colouring, a technique for passively monitoring network traffic.