Optical burst switching (OBS), which takes advantage of both high capacity of optical fiber and sophisticated control of electronics, has been considered as a promising paradigm for the modern Internet. In this paper, the recovery procedures for IP over OBS/WDM multilayer network are addressed, which focus on the failure detection and failure localization. Bidirectional forwarding detection (BFD) mechanism which provides low-overhead, short-duration detection of failures in the path between adjacent forwarding engines, including the interfaces, data link(s) is introduced for the failure detection and localization in the IP over OBS/WDM multilayer network, especially at the OBS/WDM layer. The unique feature of OBS and the easy-to-implement merit of BFD make it feasible for BFD adopted at OBS/WDM layer. Besides, BFD is a highly cost effective method when compared with traditional methods such as detecting the Loss of Light of each wavelength. Not only the implementation details of BFD-based failure detection for the OBS/WDM layer are demonstrated in this paper, but also how the BFD localizes failures such as fiber failure, link failure and node failure. Moreover, extensive simulations have been done at different scenarios such as single failure and multiple failures. And the numerical results concerning the performance of BFD-based failure detection and localization in terms of failure detection and localization time and BDP lost rate verified the proposed solution.typical features of OBS network, such as separation of control channels and data channels as well as one-way reservation protocol, make it difficult to directly adopt the protection or restoration techniques devised for the other WDM-based optical networks [8][9][10][11][12][13][14]. In this paper, the recovery procedures for IP over OBS/WDM multilayer network are addressed, which focus on the failure detection and failure localization. The failure detection and failure localization play an important role in the process of protection or restoration for optical networks in that shorter failure detection and quicker failure localization mean less loss of data (significantly meaning for the optical networks whose data transmission rates are always up to Tbit/s per fiber) and less restoration delay (important for those delay-sensitive applications). In IP over OBS/WDM multilayer networks, failure detection and failure localization can be implemented either by the IP layer or by the OBS/WDM layer. In the first case, MPLS-based techniques such as MPLS LSP Ping and MPLS LSP Trace route can be adopted [15]. Obviously, it takes too much time to be interesting from the viewpoint of automatic recovery methods. In the second case, faults can be detected by Loss of Light (LOL) [8] or the control plane mechanism such as Link Management Protocol (LMP) based on GMPLS protocol [16,17] and similar messages like 'Keep-alive' control message [11] and 'Link-liveness' message [13]. Thus, the failure detection and failure localization in the OBS/WDM layer is quicker compared w...