Optimistic Fast Rerouting
Hai-Khun Tan, Tung-Wei Kuo · 2022
Due to the increasing number of switching devices, failures regularly occur in modern computer networks. When a packet encounters a failed link, it is forwarded along a failover route. In fast rerouting, failover routes can be established quickly based on the pre-installed forwarding entries without the involvement of the control plane. Most prior research on fast rerouting sacrifices the quality of failover routes (e.g., the length of the failover route) for packet delivery guarantee. In this paper, we propose a fast rerouting framework that consists of two forwarding modes, the optimistic mode and the fallback mode. After a packet encounters the first failed link, it is forwarded in the optimistic mode, whose goal is solely to optimize the quality of failover routes. If the optimistic mode fails to find a high-quality failover route, the proposed framework then adopts the fallback mode, which can employ any fast rerouting algorithm that guarantees packet delivery. The simulation results show that the proposed framework can significantly shorten the failover routes and thus reduce the end-to-end delay. We believe that the proposed framework can also optimize other performance metrics related to failover routes.