Abstract
Most networks are inherently prone to failures, and failures do occur on a regular basis. New real-time services, relying on continuous connectivity, are regularly introduced on the Internet - requiring new demands to be met, often extending beyond the Internet's original design goals. Traditionally, recovery has been covered by the IP re-convergence process, which is a lengthy process offering recovery time in the range of seconds. In this thesis IP Redundant Trees (IPRT), a new method for providing IP fast recovery, is introduced. It is based on the redundant tree approach presented by Medard et.al, extended to provide a resource efficient way to populate the recovery Forward Information Base (FIB) and furthermore, the mechanisms needed to utilize this information in the forwarding procedure in order to provide local recovery. IPRT is evaluated through the use of graph theory in the initial design phases, and simulations on several real and synthetically generated networks. The evaluation shows that one of the strongest assets of IPRT is the ability to provide 100 \% coverage with a minimal and constant amount of extra state information in each router.