Computing distance-bounded node-disjoint paths for all pairs of nodes — An application to optical core network design

Core optical networks require connectivity between all pairs of nodes. If the optical core network is transparent then the length of fibre between each pair of core nodes must be within a given threshold. A failure in the network could affect hundreds of thousands of customers. Therefore, it is important to ensure multiple connectivity between all pairs of nodes through at least two node-disjoint paths while respecting the path-length constraints. Node-disjointness allows switching to an alternative path whenever a link or a node fails. Minimising the cost associated with the fibre while maximising the disjointness for a given network is a very challenging task. We model this problem and develop an approach that is scalable both in terms of time and space for solving very large-sized problem instances.