Many carriers are migrating away from SONET ring restoration toward mesh restoration through intelligent O-E-O cross-connects (XC). In mesh restoration, when failure occurs, traffic along the service path needs to be quickly rerouted to the restoration path. Due to the nature of distributed protocols, several restoration paths may compete for limited capacity on the same link, which
The reason for crank back can either be different paths competing for the same channel, which we call glare, or there may not be enough free bandwidth available at all. Our research aims to develop mechanisms to reduce these two sources of crank back. We developed channel selection methods  that considerably reduced glare, while at the same time maintained large groups of contiguous channels (for high bandwidth connections) across multiple parallel links. In addition, we proposed and analyzed a hybrid solution  that utilized a centralized restoration path server to optimize the restoration path selection, yet utilized distributed control to compute service paths and set up service/restoration paths. Our objective was to achieve very fast provisioning, fast restoration upon network failure, and efficient use of capacity. To support this, we also presented an efficient restoration path selection algorithm.