Abstract: Internet is increasingly becoming a critical information infrastructure where even transient failure causes undesirable damages. BGP, the de facto standard routing protocol, is well known to suffer slow and reactive convergence process to follow the dynamic changes in the network topology and policy changes such as link or node failures. Many measurements show that BGP experiences the frequent loss of connectivity and routing loop during the routing re-convergence periods, which degrades the forwarding performance of data plane and so it is hard to support mission critical application. In this paper, we present a lightweight BAR-BGP protocol for achieving fast recovery from the failures of interdomain link failure, which deals with transient failure by combining the method of backup AS advertisement and reactive re-computing feasible paths. BAR-BGP provides a feasible relay forwarding of AS node which can be used to encapsulate packets to relay AS node immediately in case of AS node experiencing transient failure through advertising backup AS with very low message and storage overhead. It uses “root cause notification” to prevent routing loops during convergence. Furthermore, BAR-BGP uses recovery forwarding mode to improve the loss of connectivity in case that backup AS addresses are not available. Through the detailed simulation on Internet-like topology, we suggest that BAR-BGP improves Internet reliability, reduces the transient failure and forwarding disruption effectively during the routing convergence periods.