In this paper, we propose a fast and reliable fault recovery algorithm, named a resource look-ahead algorithm, for Multiprotocol Label Switch (MPLS) network. The proposed algorithm can alleviate the notification time according to the number of hops, and improve the mean fault recovery time. The algorithm uses a resource look-ahead table, which contains pre-reserved resource information. With using this table, a General Switch Management Protocol (GSMP) controller manages switches included in a given domain of open network architecture. Thus, the propagation delay of reservation message delivery can be alleviated. The restoration time is also improved comparing to the MPLS-based fault recovery.