Allocation and Fragmentation Policies for Routing and Spectrum Assignment in Elastic Optical Networks

The exponential growth of heterogeneous, bandwidth-intensive applications has revealed the limits of traditional fixed-grid Wavelength Division Multiplexing (WDM) networks and motivated a shift to flexible-grid Elastic Optical Networks (EONs). EONs offer much higher spectral efficiency but also create a challenging combinatorial problem: Routing and Spectrum Assignment (RSA). RSA must respect spectrum continuity and contiguity constraints while avoiding fragmentation in a dynamic traffic environment. This thesis offers a comprehensive study of RSA strategies for EONs through a purpose-built discrete-event simulator. The simulator models dynamic lightpath setup and teardown under heterogeneous traffic conditions and implements key RSA policies — First-Fit, Random-Fit, and Longest-Fit — as well as important architectural options such as wavelength conversion and a software-defined break capability. Experiments were carried out on both toy and large-scale realistic topologies and validated against reference implementations. Our results show that deterministic allocation policies generally outperform stochastic ones. First-Fit minimizes blocking probability, while Longest-Fit maximizes resource efficiency: it reduces wavelength converter usage by 18–24\% while keeping blocking performance competitive. Enabling break capability proved to be a highly effective, low-cost software feature that significantly improves performance and reduces hardware requirements. By contrast, wavelength conversion yields only modest blocking reductions (2–6\%) at substantial economic cost. Based on these findings, we recommend Longest-Fit with break capability for resource-conscious deployments and First-Fit where minimizing blocking is the priority. This work contributes a validated simulation tool and actionable insights for RSA design, laying the groundwork for future research into machine-learning driven, impairment-aware, and cost-aware optimization of elastic optical networks.