Abstract. Large IP backbone networks today are mostly deployed directly over sequences of point-to-point DWDM systems or chains of newer ROADM-based ultra long haul systems, interconnected by OEO regenerators. The next generation core optical network is moving toward an all-optical network architecture that is based on multi-degree ROADMs to reduce OEO regeneration cost as well as enabling automatic reconfigurability and dynamic restoration. In this paper, we study the restoration design in this new IP over reconfigurable all-optical network architecture to satisfy the resilience requirements for both IP and wavelength services. We propose two novel restoration schemes: 2-Phase Fast Reroute mechanism with optimized Traffic Engineering algorithm for restoring IP services and shared mesh restoration with standbys for restoring wavelength services. They both meet the requirement of sub-second restoration time and also maximize sharing among different failures with the objective of minimizing either overall capacity or overall cost. To further reduce the required restoration capacity in both IP layer and optical layer and address failures in both layers efficiently, we also propose an integrated IP-over-optical layer restoration strategy that enables sharing of restoration capacity among non-simultaneous failures across both IP and optical layers. Simulation results demonstrate significant improvements using our proposed schemes comparing with existing ones.