Resource Allocation in Flexible-Grid Optical Networks with Nonlinear Interference

Licentiate thesis, 2015

As the backbone of modern communications, the optical networks are anticipated to provide higher data rate and flexibility to support the exponentially growing volume and heterogeneity of traffic requirements. Flexible-grid optical networks have been proposed to improve the utilization of spectrum resources. However, the physical layer conditions are more complex and variable in flexible-grid networks than fixed-grid wavelength division multiplexing networks. Therefore, the consideration of physical layer impairment (PLI) is necessary in the planning stage of flexible-grid optical networks. In this thesis, we mainly study the allocation of physical layer resources such as modulation formats, power spectral densities (PSDs), and carrier frequencies for all the channels in flexible-grid networks. To accurately estimate the quality of transmission, both linear and nonlinear PLIs are considered. Novel resource allocation problems are formulated and solved to include a nonlinear channel model and additional degrees of freedom in flexible-grid optical networks. The network performance and efficiency improvements over previous works are demonstrated via numerical calculations. Paper A proposes a resource allocation algorithm for a single flexible-grid fiber link based on a nonlinear signal distortion model, i.e., the Gaussian noise (GN) model, as a first step towards a whole-network design. Based on the accurate estimation of channel PLIs provided by the GN model, the proposed algorithm can allocate resources more efficiently in terms of spectrum usage. Its performance is demonstrated through comparisons with the benchmark algorithm utilizing the transmission reach method. Paper B and Paper C extend the proposed formulation to the network level and study the impact of nonlinear interference on transmission reaches and resource usage. Specifically, with precalculated routes and spectrum orderings for all the channels, Paper B allocates resources in a three-node network with optimized PSD. In Paper C, the algorithm is further extended to more complex networks to demonstrate its scalability and performance. In Paper D, we jointly allocate route, spectrum, modulation format, and PSD for each connection request in the flexible-grid network. A mixed integer linear programming problem is formulated to search for the optimal resource allocation. A heuristic algorithm based on problem decomposition is also developed to reduce computational complexities. The joint resource allocation approach can improve the spectrum efficiency even further compared with previous separate planning strategies.