Wavelength Switching in Software-defined Optical Transmission System

The continuous growth of Internet traffic, including HD video rendering, cloud computing, and the Internet of things (IoT), motivates more efficient optical communications capable of handling a wide range of applications. Dynamic reconfigurable optical add-drop multiplexer (ROADM) based wavelength-division multiplexing (WDM) systems in which connections are established through real-time wavelength switching operations have long been studied as a means of achieving greater scalability and increasing the network resource utilization. While ROADMs are extensively deployed in today's WDM systems, they remain quasi-static with wavelengths being provisioned to meet the peak traffic requirements and left in place. Recent advances in hardware and software greatly improve the multi-layer control and management of ROADM systems facilitating wavelength switching. However, ensuring stable performance and reliable quality of transmission (QoT) remain difficult problems for dynamic wavelength switching. A key challenge in converting today's quasi-static ROADM systems into real-time wavelength routed ROADM systems is the optical power dynamics that arise from a variety of physical effects in the amplifiers and transmission fiber.;In this dissertation, we first discuss the model of the power excursion due to interactions between the wavelength dependent gain and automatic gain control of optical amplifiers. The power excursion is the main manifestation of optical power dynamics. In order to offer rapid wavelength switching capabilities for dynamic ROADM transmission systems, different hardware and software methods are investigated to mitigate power excursions. First, a dual-wavelength source is implemented to distribute a single optical signal over two wavelengths---one with a high gain and the other with a low gain---to equalize the average gain and mitigate power excursions. Second, we investigate non-disruptive and proactive probe methods to predict a variety of physical parameters and thus recommend optimal wavelength assignments with minimal power excursions. Third, a deep neural network based machine learning method is investigated to predict the optical power dynamics from data collection and training. The trained deep neural network can recommend wavelength assignments for wavelength switching with minimal power excursions.;Software-defined networking (SDN) is a key to providing software control capabilities that can be exploited to achieve real-time wavelength switching in large-scale multi-domain ROADM systems, but its scalability and flexibility are limited by various types of physical layer impairments. In this dissertation, we investigate a transparent software-defined exchange (TSDX) control system to guarantee service level agreement (SLA) requirements for multi-domain optical systems in which optical signals can be exchanged between domains entirely in the optical layer through Internet exchange points (IXPs). Advanced optical performance monitors (OPM) are implemented for real-time introspection of the physical system at different network locations without the need for optical-to-electrical-to-optical (OEO) processing. Real-time optical signal to noise ratio (OSNRs) measurements using OPMs together with inter-domain negotiation enabled by the SDN-based hierarchical control architecture allow impairment-aware wavelength rerouting and code adaptation in our 6-ROADM multi-domain optical network testbed.