Abstract. In this paper we present a policy-based architecture for aggregating (grooming) IP/MPLS flows (packet-based LSPs) within lightpaths taking into account the possibility of having to cope with further transport faults. The defined policies try to minimize the negative impact when a failure is detected in the optical transport network. Such policies deal with 1+1, 1:1 and 1:N schemes of protection. In our model, IP/MPLS flows are divided into High Priority (HP) and Low Priority (LP) traffics. The architecture is composed of an Admission Control responsible for receiving the requisitions from the IP/MPLS network and forward them to the Policy Manager which in turn is responsible for applying the policies. The architecture also has a Fault Manager responsible for accounting the failures and a Resource Manager responsible for managing the lightpaths. Our approach has been implemented to validate the policies and the results showed that the defined policies decrease the number o...