The strict Quality of Service requirements of real-time services in communication networks necessitate the fast recovery from link/node failures by activating alternate paths which avoid the failed component. Multi Topology Routing (MTR) based IP Fast Re-Route (IPFRR) technologies rely on virtual topologies (VTs) to seamlessly forward the IP packets during network failures. In this paper, we implement a new MTR based IPFRR technique for Software Defined Networks (SDN) which uses the Multiple Routing Configurations method to construct VTs. Our approach leads to self-recovering SDN against failures. Out preliminary results show that our approach considerably reduces the failure recovery time compared to the reactive recovery process in SDN.

To seamlessly support real-time services such as voice and video over next generation IP networks, routers must continue their forwarding tasks in case of link/node failures by limiting the service disruption time to sub-100 ms. IETF Routing Area Working Group (RTGWG) has been working on standardizing IP Fast Reroute (IPFRR) methods with a complete alternate path coverage. In this paper, a trade-off analysis of Multi Topology Routing (MTR) based IPFRR technologies targeting full coverage, namely Multiple Routing Configurations (MRC) and Maximally Redundant Trees (MRT), are presented. We implemented a comprehensive analysis tool to evaluate the performance of MRC and MRT mechanisms on various synthetic network topologies. The performance results show that MRT's alternative path lengths are not scalable with respect to the network size and density while the alternative path lengths of MRC only slightly change as the network size and density vary. We believe that this is an important scalability result providing a guidance in the selection of MTR-based IPFRR mechanism for improving the availability in ISP networks.