Impacts of Physical Layer Impairments on Optical Network Performance
Nowadays, dynamic traffic demands --- including high-definition video streaming requested by mobile devices and large bulks of data transfers taking place between data centers --- continue to increase in volume over time. The present networks are struggling to meet the dramatic data rate increase over the next decade. Optical networks form the backbone of modern communication systems and are evolving to be able to offer transmission rates as high as 1~Tbps per channel (wavelength). Since the usable optical spectrum is limited, how to increase the bandwidth efficiency to accommodate all the requests while considering the transmission impairments that degrade the optical signal quality is also a big challenge. With network traffic being more dynamic, one important issue is the signal quality fluctuation over the time when other optical circuits (lightpaths) are set up and torn down. The quality of transmission (QoT) must be satisfied and ensured during the entire lifetime of any lightpath. Physical layer models are used to evaluate the signal quality. These models need to be efficient in order to estimate the QoT in real time, considering the current network status. In this dissertation, several models for estimating the physical layer impairments (PLIs) and the signal quality with the feasibility to deal with dynamic traffic are analyzed. Network layer solutions and algorithms are proposed to improve the QoT. Three models and their applications are studied: 1) legacy models with non-coherent detection in extended Generalized Multi-Protocol Label Switching (GMPLS) based WDM networks, in which race conditions from both normal traffic and restoration requests when fiber link failures happen are analyzed and efficiently used to improve the acceptance rate; 2) a modeling framework to estimate signal power and channel optical signal-to-noise ratio (OSNR) accounting for realistic characterization, i.e., uneven gain and noise figure, of optical amplifiers is introduced and its accuracy and efficiency in both line networks and meshed networks are analyzed; 3) Gaussian Noise (GN) model in transmission systems with coherent detection and its application with spectrum assignment constraints to improve channel OSNR in GMPLS based networks are studied. With the above models, network operators would be able to accurately estimate the signal power and channel OSNR to ensure the QoT of all the channels is well maintained within the receiver's sensitivity region. Moreover, the improved acceptance rate and signal quality can significantly increase the network throughput and bandwidth efficiency.