The Internet is nowadays used for a diverse set of services. Some of them have stringent demands regarding latency, data rate, jitter, etc. Some examples of applications which make such demands are Voice over IP, IP TV, telemedicine, stock exchange information, gaming, etc. In general the diversity of tasks being solved by using the Internet as the communication medium continuously increases the demand availability of connections which is the focus of this thesis.
This thesis focuses on discussing and evaluating recovery schemes that fit the IP Fast Reroute Framework (IPFRR) developed by the IETF Routing Area Working Group. IPFRR is a good framework for recovery schemes that provide fast reroute in connectionless IP networks. Two recovery schemes have been evaluated, namely IP Fast Reroute Using Not-via Addresses (Not-via) and Failure Insensitive Routing (FIR).
The concept of the Not-via recovery scheme is that whenever a link fails, it will consider the neighboring node as down. It will then find a path towards the next-next-hop towards the destination from the path in the failure free case.
The FIR scheme on the other hand is able to infer network failures by looking at the flight of a packet. The flight of a packet refers to the path it takes through the network. Based on this the scheme is able to proactively create forwarding tables that makes sure that traffic never traverses failed network elements.
To be able to compare performance of the recovery schemes in both real life and synthetic networks, a routing simulator was developed. It was used to show that for most networks, FIR will provide shorter recovery paths than Not-via. Since longer recovery paths leads to more links beeing used for recovery traffic, Not-via will introduce a larger amount of load than FIR. However, an important feature of the Not-via scheme is that it is able to recover from both node and link failures, whereas the FIR scheme is only able to recover from link failures.
It has found that IP Fast Reroute using Not-via addresses would probably be a better choice if a scheme should be implemented in hardware since it has less requirements for doing so. This scheme also has better coverage than the Failure Insensitive Routing scheme and as such would be a better choice for any network operator.