Switching Architectures And Wavelength/Waveband Routing In Multi-granularity Optical Networks

With the development of dense wavelength division multiplexing (DWDM) technology, the number of the wavelengths per fiber is becoming large. UP to now, a DWDM system with 160 wavelength channels has been used commercially and a system multiplexing 1022 wavelength channels has been completed experimentally. Multi-granularity optical node is introduced as an effective, economic and scalable method to manage a large number of wavelengths in DWDM optical networks. As the key element in optical networks, switching architecture is a very important research field. In addition, effective routing and wavelength/waveband (RWA) may realize effective switching resources and bandwidth resources assignment to achieve better switching interconnection and better networking performance. Therefore, RWA is one of the most important problems in multi-granularity optical networks.This dissertation focuses on the multi-granularity switching architectures and RWA in multi-granularity optical networks. The contents of this dissertation are summarized as follows.The first chapter introduces the related background.Two novel multi-granularity optical cross-connects (MG-OXC) architectures including the interconnecting and the parallel MG-OXCs are proposed in Chapter 2. Several MG-OXC architectures are studied and analyzed on physical performances. Then the comparison is carried out in terms of the number of ports and the blocking performance. In addition, the interconnecting three-stage MG-OXC with express paths is expanded to n-stage interconnecting MG-OXC, in which n granularities switching fabrics are interconnected by adaptive modules according to the optimized proportion. Compared with the serial and single layer MG-OXCs, the proposed MG-OXCs are flexible for dynamic traffic and can significantly reduce optical losses, occupied port count and cascaded filtering penalties which could translate into the network cost and size. Finally, the drop functionalities of the serial three-stage MG-OXC and interconnecting three-stage MG-OXC are demonstrated in 10 Gbit/s WDM system by the experiment.In Chapter 3, the cross-connects configuration of the MG-OXC is investigated based on the dynamic traffic model in all-optical networks. Firstly, a bipartite graph G (X, Y, E) with Boolean array is used to describe three-stage MG-OXC architecture. Then, we propose a configuration algorithm based on the bipartite graph G (X, Y, E) to satisfy the dynamic arriving requests with different bandwidth. In addition, the simulation and the wavelength grooming experiment on the MOST (Multi-functional optical switching testbed,MOST) are carried out to evaluate and validate the performance of configuration algorithm. Finally, the associated architecture parameters in the serial three-stage MG-OXC and system parameters in the system are analyzed in detail and optimized by simulations. The proposed optimization design significantly reduces the node complexity and improves the blocking performance of arriving requests, which translates to smaller device size and lower cost in optical networks.Chapter 4 proposes an on-line RWA algorithm in DWDM multi-granularity optical networks and analyzes the performances of multi-granularity networks based on the serial MG-OXC and interconnecting MG-OXC. Firstly, a layered waveband graph, bundling requests scheme and routing strategy are presented in detail. Then, we propose an on-line RWA heuristic algorithm, called Maximize the Number of Waveband Route (MNWR), for the dynamic arriving traffic demands in the multi-granularity waveband switching optical network. It is clear that the MNWR algorithm based on waveband route can result in high bandwidth utilization and ports utilization, and hence the better blocking performance. Next, we study the performances of various MG-OXC architectures including the serial three-stage MG-OXC and the interconnecting three-stage MG-OXC employing the on-line MNWR algorithm. Simulation results indicate that the interconnecting three-stage MG-OXC outperforms the serial three-stage architecture in terms of blocking performance for dynamic requests due to its integrity and flexibility, which translates into the lower network cost. Lastly, we study some parameters on the multi-granularity optical networks by simulation. The results show that different parameters have different effects on the serial and interconnecting MG-OXCs network. In addition, the optimized parameters can improve the network performance significantly and reduce the network cost.The RWA issue of ring networks employing novel parallel multi-granularity hierarchical optical add-drop multiplexers (MG-OADMs) is investigated for the off-line case in Chapter 5. Firstly, we introduce the parallel MG-OADM architecture based on blocker filters and the related research. The objective of RWA problem is to minimize the number of control elements in the off-line network based on parallel MG-OADMs. Then, we present an integer linear programming (ILP) formulation to obtain the lower bound in optimization. However, for large-scale OADM networks, the ILP approach would be prohibitively time-consuming thus efficient heuristics are desired. Therefore, we propose an efficient heuristic algorithm called Global Bandwidth Resource Assignment (GBRA) that is suitable for the design of large-scale OADM networks. The GBRA algorithm minimizes the number of fragments in each waveband as well as the number of nonempty wavebands by an efficient global assignment strategy utilizing existing nonempty wavebands which could lead to the fewer control elements, thus achieving a close-to-optimal solution. In addition, the complexity of the GBRA algorithm is equal to that of the Node Assignment First (NAF) algorithm proposed by Y. Su etc. Finally, we investigate some scenarios of different scale networks with different traffic distributions and show the numerical results. The results show that the GBRA algorithm is more efficient than the NAF algorithm for optimization.