Reactive Load Balancing During Failure State in IP Fast Reroute Schemes

Abstract

To improve reliability of IP networks against link/node failure, several IP fast reroute schemes have been proposed so far. They proactively compute backup paths and activate them when failure occurs to prevent packets from losing at the failure link/node. However, it is known that network performance considerably degrades in the failure state of IP fast reroute schemes, because congestion hot spots often appear near the failure link/node. In this paper, we propose a reactive load balancing method that can be applied to the major IP fast reroute schemes that covers single failure. Our scheme works when an IP fast reroute scheme is activating its backup paths, and reduces the degradation of network performance due to failure. In our load balancing scheme, with the overhead of a few bit field on packet header, we can largely reduce the performance degradation in the failure state and mostly keep the throughput as it was in the normal state where no failure exists.