CAM05-4: Handling Multiple Network Failures through Interface Specific Forwarding

It has been observed that transient failures are fairly common in IP backbone networks and there have been several proposals based on local rerouting to provide high network availability despite transient failures. Previously, we proposed failure inferencing based fast rerouting for IP backbone networks that ensures delivery of a packet to its destination if there exists a path when a single link fails but can cause forwarding loops in case of multiple simultaneous failures. On the other hand, blacklist-aided forwarding, we proposed earlier for wireless mesh networks, provides loop-free forwarding even in the presence of multiple failed links in the network but requires that each packet carry a blacklist of failed links encountered along its path. Our aim is to achieve the best of both these approaches, i.e., successfully deliver packets while ensuring loop-freedom even in case of multiple failures without changing packet format. We propose blacklist-based interface-specific forwarding (BISF) that infers a blacklist, a list of links that might have failed, based on a packet's incoming interface and its destination, and determines the next-hop by excluding the blacklisted links. We show that BISF is loop-free regardless of the number of failures in the network while forwarding packets successfully in most cases.