| Along with the development of WDM technology, a lightpath maybe carry lots of data steams. A single fiber failure of fiber can lead to the failure of all the lightpaths traversing the fiber, and results in significant loss. Hence, it is necessary to design fault-tolerant optical network. All kinds of protection technologies are good for fault torlerance. Service has been payed attention to the optical networks, i.e., ensuring QoS, providing differentiated service has becomed important object of designing fault-tolerent optical networks. Availability is an important metric to measure service in optical networks, so this thesis considers not only designing protection but also availability-aware utilization of spare resource in optical networks. This thesis studies availability-aware protection algorithm,"working-protection"sharing, multicast protection, availability-aware wavelength assignment.The availability-based routing algorithm is studied. The metric of reliability has ever been considered to provide the differentiated services, but it is the failure probability in certain period of time, which only denotes the"hit"probability of the fault. However, in the dynamic networks, all the traffic comes or leaves dynamically and the repair time of the fault should be considered, so the availability is better to scale optical networks than the reliability. A layered structure to analyze the network availability is given. By using the structure, this thesis analyzes the availability of primary and backup links separately. Through the simulations, the successful route ratio of our algorithm shows better than others'.The wavelength assignment is normally realized through the First-Fit algorithm. The differentiated service should be satisfied in the aspect of the wavelength assignment, so the thesis gives an availability-aware wavelength assignment algorithm, which limits the scope of wavelength selection of shorter lightpaths. Limiting wavelength can be performed through a wavelength threshold that the maxim wavelength is multiplied by the ratio between the requested and offered availability. Through this type of wavelength assignment, the longer lightpath has more chance to select the available wavelengths. In simulations, the algorithm limiting wavelengths of the shorter lightpath can make the average availability and blocking performance enhanced. The blocking of the longer lightpath is a main factor to affect the overall blocking performance. Increasing the chance of wavelength selection of the longer lightpaths decreases the blocking probability of the longer lightpaths. Otherwise, the algorithm can show better in the large-scale networks, because the proportion of the longer lightpaths is larger correspondingly in this type of networks. Improving the performance of the longer lightpaths may bring obvious gain.In order to enhance sharing effiency of"working-protection", three patterns have been defined for the protection links: preemption, orphan and mix, which relate to the requested availability of incoming traffic. The different patterns can lead to the different availabilities. The simulations show that the blocking performance and the average gain of availability are all enhanced while considering the patterns of spare resources. Furthermore, the thesis studies on multicast protection. IfSRLG(sharing risk link group) is considered, the spare resources of in-tree and out-tree are limited. In order to impair the influence of SRLG, the proposed algorithm adjusts the sharing degree of spare resources. A novel link cost function is designed so that the spare resources are shared effectively. Through simulations, the overall successful session ratio of proposed algorithm is better than the algorithm not considering the sharing degree between in-tree and out-tree. The proposed algorithm is not sensitive to the span-out of trees. Along with the increase of span-out, while encouraging the in-tree sharing, many protections can be found in the primary light-trees.
|
PDF Full Text Request