On the Feasibility and Efficacy of Protection Routing in IP Networks

With network components increasingly reliable, routing is playing an ever greater role in determining network reliability. This has spurred much activity in improving routing stability and reaction to failures, and rekindled interest in centralized routing solutions, at least within a single routing domain. Centralizing decisions eliminates uncertainty and many inconsistencies, and offers added flexibility in computing routes that meet different criteria. However, it also introduces new challenges; especially in reacting to failures where centralization can increase latency. This paper leverages the flexibility afforded by centralized routing to address these challenges. Specifically, we explore when and how standby backup forwarding options can be activated, while waiting for an update from the centralized server after the failure of an individual component (link or node). We provide analytical insight into the feasibility of such backups as a function of network structure, and quantify their computational complexity. We also develop an efficient heuristic reconciling protectability and performance, and demonstrate its effectiveness in a broad range of scenarios. The results should facilitate deployments of centralized routing solutions. I. INTRODUCTION Intra-domain routing in IP networks has traditionally relied on distributed computations among routers, with the concate- nation of individual forwarding decisions eventually resulting in packet delivery. In spite of their inherent adaptability and scalability, distributed computations can make troubleshooting harder, because of the many sources of inconsistencies they allow. This has renewed interest in centralized routing solu- tions (4), (6), (21) for IP networks, at least in settings where scalability is less of a concern, e.g., intra-domain routing. Centralizing decisions not only guarantees full visibility into the forwarding state of individual routers (now essentially cheap forwarding engines or FEs), it also affords added flexibility in computing paths that meet different requirements. In spite of its advantages and even when scalability is not an issue, centralizing decisions has disadvantages. Of particular concern for reliability is latency in reacting to failures, i.e., the central server needs to be notified, react to the failure, and communicate updated forwarding information to all affected FEs. This can result in non-negligible "gaps" after failures, during which FEs have no valid forwarding states for some destinations, and translate into substantial packet losses. A natural approach to the problem is through preventive mechanisms, e.g., by having the central server pre-compute