Improve reliability of scalable VPN routing via Relaying

Layer 3 MPLS VPN service is becoming more and more popular because it can provide direct any-to-any reachability for enterprise customers´ geographically disparate sites over public network. However, this model of direct reachability makes the routing table in service provider´s PE routers grow very fast. C. Kim et al. propose Relaying [1], a scalable VPN routing architecture, to solve this routing table explosion problem in PE routers. They divide the PE routers in a VPN into two groups: hubs and spokes. A small number of hubs maintain full reachability information for that VPN, while spokes with reduced reachability information achieve any-to-any reachability by delivering traffic via hubs. However, one important concern about Relaying is the potential decrease of availability due to the small number of hubs. Since a hub router is critical to maintaining reachability between its spoke PEs, when one of the hubs in a VPN goes down, the spokes who route traffic via this hub can´t reach each other. In this paper, we propose a backup mechanism to improve the reliability of Relaying by assigning each spoke to one main hub and another backup hub. We also propose an implementation technique by simply modifying the configuration of routers so that the spokes can automatically switch to backup hub when the main hub goes down and automatically switch back whenever it is repaired. The simulation result shows that our backup mechanism can greatly improve the reliability of Relaying with small performance penalty.