Optimal Resource Provisioning for Dynamic Wavelength Services With Access Port Constraints

With increasing demands for dynamic high data-rate private line services from large customers, network carriers have begun to provide flexible on-demand connections to construct these customers' networks. Advances in optical devices have greatly improved the reconfigurability of the optical layer, which makes it possible to provide optical on-demand services at wavelength granularity (10 Gbps to 40 Gbps per channel). This new dynamic wavelength service model allows a customer owning or leasing a few terminal ports in the network to connect these ports on demand. The network must be pre-dimensioned with enough resources to support any possible connection configuration between the pre-allocated ports of a dynamic wavelength customer. In this paper, we introduce a resource optimization problem for dynamic wavelength services and propose a few efficient techniques to tackle this challenging problem. A heuristic optimizer based on a combination of simulated annealing and genetic algorithms is developed to solve the problem efficiently for large networks. We show that our solver can produce optimization solutions within 7% overhead of a lower bound with a small amount of computation.