NeTS: Small: Cognitive Management and Control of Agile Dynamic Optical Networks

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

Part 1. The optical network of the future will have orders of magnitude increase in data rates, due at least partially to the increase in big-data transactions. These create the need for fast scheduling of network resources and agile network adaptation to most efficiently move the data across the network. This project proposes to investigate a cognitive network management and control system, which 'senses' current network conditions and uses this information to satisfy overall performance goals. This project will be the first comprehensive research on cognitive optical network management and control. The goal is for agile automated adaptation to replace current slow, manually-driven management and control practices. The fruits of this research will have implications for next generation wireless networks and power grid systems and for fast detection of extreme events that can significantly disrupt networks. Part 2. Current optical networks are operated with predominantly static connections, with lightpaths changed quasi-statically and often remaining unchanged for months. Present methods of setting up a wavelength path result in slow changes to the network (~20 min setup times), as each of the network elements along the path is gradually tuned to avoid instabilities and transient impairments arising from rapidly introducing another optical channel into the network. Large flow and other traffic dynamics will require bandwidth adaptations of the order of ~100mS-1S. In today's optical networks, the link quality of all adjacent wavelengths is monitored as the lightpath is turned on in several steps. The number of network parameters and their short coherence time in dynamic future networks render acquiring the complete state information of the network impractical. The idea of this project is to sense and control a small subset of the parameters with the 'largest' information content and use their relationship to tune the network. Cognitive techniques that need to be incorporated in such a strategy include sampling of network and link parameters, applying data analytics to learn behavior and cognitive techniques to operate networks, and inferring network state and optimizing performance without placing an undue network control and management communications burden on the network. This research will consider both dynamic per-session scheduling, rerouting/load-balancing and topology reconfiguration to optimize network performance.