Future communication networks will carry many WDM channels at very high bit rate. Thus, it is desirable to avoid electronic switching at the core. The all-optical backbone network can be interconnected by optical cross-connects at strategic locations to allow for flexible capacity provisioning and fault-tolerant rerouting. Such an all-optical core layer nicely decouples the long-term capacity planning problem from the short-term dynamic bandwidth allocation problem which can be better tackled in the electronic domain. An essential requirement for the all-optical core layer is that it must be fully fault-tolerant, otherwise, a single failed link can cause a disaster for the entire network.
We consider the problem of how to allocate the required capacities and spare capacities on a given all-optical core network so as to make the network fully single-fault (or multi-faults) tolerant. The objective function is the total cost of the spare fibers. Based on a given traffic requirement on all source-destination pairs, the optimal bandwidth requirement for each link in the given topology is first computed. We then consider link failures one-by-one for the entire network. For each link failure, we show how spare capacity can be added in other links so as to take advantage of existing spare capacities that have already been added. The algorithm is based on the shortest path routing algorithm and has a polynomial time complexity. Preliminary investigations suggest that the algorithm can give results comparable to those obtained by integer programming.