Society has shown an endless appetite for new data-rich applications running over the Internet, which depends on fiber-optic networks to carry its traffic. Elastic optical networks (EONs) are widely seen as the logical evolution of current large-scale fiber-optic core transport networks due to their flexibility and ability to support growing capacity demands. They can also deploy advanced signal processing techniques to balance the requirements of high resource efficiency and high signal quality. The goal of this proposed work is to design long-haul EONs and invent algorithms that wisely employ its system resources, including the optical spectrum as well as physical network equipment. The objective is to minimize resource usage, implying low capital expenditure and low operational cost, while satisfying expected increasing traffic needs. Better network design and dimensioning algorithms will enable providers to meet their customers demands far more quickly and cost-effectively.
The research entails using an accurate model of physical-layer impairments within the design and management of translucent EONs. By using a network-state-dependent performance prediction instead of the transmission reach in the routing and resource allocation procedure, significant spectral and equipment savings can be gained. This cross-layer approach trades off computational complexity for increased resource usage efficiency. The research agenda begins by modifying the currently popular Gaussian-noise model for physical layer impairments for modern long-haul networks to be easily computed as a function of the network state. Then the project addresses the design of long-haul EONs by solving two fundamental problems: - Algorithm design for routing and resource allocation based on accurate physical impairment estimation; classical approaches to solving this problem have relied on worst case estimates that significantly overprovision the network. - Theoretical models for EON behavior in the presence of network-dependent physical layer impairments;the goal is to abstract physical layer parameters to make resource allocation calculations more tractable and less sensitive to small changes in those parameters. The project will be performed in collaboration with a research team at Nokia Bell Labs to ensure industrial relevance of the work.
