A Fast Failure Recovery Scheme for Fibbing Networks

Fibbing is a network technology that can provide flexible routing in IP networks. In a Fibbing network, the network controller cleverly broadcasts link state advertisements (LSAs) of the open shortest path first (OSPF) protocol to generate fake nodes. These fake nodes enlarge the physical topology and turn the network into a virtual network. Although the routers still follow the shortest path routing in the virtual network, the flows are steered along the desired paths in the physical network. Even though a Fibbing network is capable of flexible routing, methods to achieve fast failure recovery have not been well studied. Conventionally, an IP network running the OSPF protocol takes a long time to converge after a failure occurs. The IETF proposed loop-free alternate (LFA) IP fast reroute technology to speed up failovers. However, we discovered that LFA technology is incompatible with Fibbing. The direct application of LFA technology in a Fibbing network will generate traffic loops. In this work, we propose a novel monitoring-cycle-based approach for fast failure recovery. By examining the liveness of the monitoring cycles, the system controller can promptly identify a failed link or a failed node and then perform traffic rerouting. Due to the properties of IP packet forwarding, a monitoring cycle in an IP network must be a simple cycle. We prove that any network that is both 2-vertex- and 3-edge-connected (2V3E) possesses a set of monitoring cycles that can be used to detect and identify any single link failure. In conjunction with the status pattern provided by the monitoring cycles, any single node failure can be detected by introducing an additional probe message. We propose an algorithm to obtain these monitoring cycles and design and implement a Fibbing controller for the network. To evaluate the performance of the proposed approach, we construct a physical testbed and an emulation network. The experimental results show that the proposed system works stably and can achieve a fast failover within a short period of time.