| Many researchers have focused on addressing a few problems of physical and virtual topology dependencies in survivable WDM optical networks and have concentrated on them separately. If physical and virtual topology dependencies are not considered in restoration, optimal restoration may not be achievable. The problem related to physical and virtual topology dependencies is NP-Hard, therefore, an efficient restoration technique is required. Additionally, restoration techniques available in literature have not addressed their performance in terms of significant, multiple objective goals. Some of these methods have shown good performance for a single objective function. However, restoration must consider a number of objective functions.; In this dissertation, we evaluate existing models and their performance in an attempt to verify their efficacy based on literature. Our research has found not only inefficiency in some of these methods of restoration, but a general incompatibility. This dissertation proposes eleven objective functions that yield objective goals significant to the optimal design of a WDM optical network. The objective goals are mathematically formulated and can be employed in the design of restoration schemes. Each objective function model is presented and is examined by experimentation. Five proposed restoration algorithms were evaluated: ASDP (Near-Optimally Annealed k-Shortest Disjoint Paths Restoration), DCROS (Deep Conjectural Reinforced Local Optimal Search), RWWA (Random Walk-based Wavelength Assignment), HRM (Hybrid Restoration Method), and PTCI (Physical Topology Connectivity Increase).; These are the major contributions (including five algorithms above) of this dissertation: Numerical results obtained by experimental evaluation of ASDP, DCROS, RWWA, HRM, and PTCI algorithms confirm that MILF (Objective Function of Minimizing Total Wavelengths with Multi-Objective Goals) based on DCROS is a technique for efficient restoration incorporating physical and virtual topology dependencies in WDM optical networks. MILF yields the best performance among other objective functions in terms of significant optimal objective goals: escape from physical and virtual topology dependencies, wavelengths, wavelength link distance, wavelength mileage costs, even distribution of traffic flows, total link and network costs, total and average restoration time of backup lightpaths required for given traffic demands. We note that experimentation was not exhaustive, yet our results encouragingly suggest the proposed methods deserve additional scrutiny. |
