Study On Path Protection In Optical Mesh Networks

In wavelength-division-multiplexing (WDM) optical networks, a large amount of traffic is sustained on the fiber. In such a network, even only one link or node failure will lead to loss of much traffic data. In WDM optical networks, network survivability has been recognized as one of the most important issues. There are two main methods to enhance network survivability, i.e., protection and restoration. Protection technology will establish backup lightpaths and reserve backup resources, while establishing working ligthpaths. When working lightpaths fail, the traffic will be switched over to their corresponding backup ligthpaths. And restoration technology will not reserve backup resources while establishing working lightpaths. When working lightpaths fail, the traffic will be rerouted and the residual link capacity has to be used. As compared with restoration, protection has a shorter recovery time and is simpler in implementation because rerouting is not needed. Furthermore, protection technology can guarantee 100% recovery, whereas restoration technology cannot. Of course, protection technology will consume more network resources. Since the optical layer lies in the bottom of network layers, it is necessary that the optical layer has a recovery mechanism with recovery guarantee and has a shorter recovery time. As a result, protection is more suitable for recovery in the optical layer. Based on whether the resources in backup paths can be shared, protection can be categorized as shared protection or dedicate protection. Compared with dedicate protection, shared protection possesses a higher efficiency of resource utilization. This dissertation mainly studies shared path protection in wavelength-routed optical mesh networks. We investigate shared path protection in both the single-link failure scenario and the double-link failure scenario. In addition, we integrate the physical topology design with protection and study the physical topology design of fault-tolerant optical networks.First, to recover from any single-link failure, we propose a novel wavelength assignment scheme. Our proposed scheme strives to utilize the partial share to improve the efficiency of wavelength utilization. Additionally, our proposed scheme can be applied in the networks with full wavelength conversion and the networks without wavelength conversion.Second, to recover from any double-link failure, we, for the first time, find that the current wavelength share scheme may lead to potential wavelength contention. We introduce the so-called preference policy. Based on this policy, we propose a novel wavelength share scheme. As compared with the current one, our scheme is simpler and eliminates the problem of the wavelength contention. Also, our scheme can provide the 100% recovery guarantee.In addition, we propose a novel method to generate the initial physical topology based on the traffic pattern for the physical topology design of wavelength-routed optical networks. The most remarkable characteristic of this method is that it makes those node pairs with more traffic have smaller distance between them. We also propose a modified routing algorithm. Unlike the traditional simple two-step routing algorithm, our proposed one takes into account the dependence of routing and wavelength share. Therefore, our algorithm has a better performance.Finally, we give a specific demonstration of the physical topology design. This demonstration takes an aim to optimize the fiber number, the amplifier number, the dispersion compensator number and the regenerator number. In addition, we adopt the taboo search algorithm to optimize the whole performance of the physical topology design. By adjusting the loop number and the taboo length, we can obtain a suitable tradeoff between the network cost and the convergence time.We solve our devised integer linear programming (ILP) formulations by using the CPLEX software. By virtual of numerical results, we obtain the following findings. The first finding is that with shared path protection, a network with only 30% nodes furnished with wavelength converters can achieve a comparable performance with a network with full wavelength conversion. The second one is that wavelength converters have a more remarkable effect on backup paths than on working ones. The third one is that with shared path protection, the cost to protect against any double-link failure is two times less than the cost to protect against any single-link failure. This indicates that recovery from the double-link failure scenario is not inefficient as we intuitively imagine.