Survivability Technologies For Self-aware Optical Switching Networks

Optical switching network based on wavelength division multiplexing technologyis at the core of the whole information networks, because it is the main artery for theend-to-end transmission of network service, which shouldering the important task ofhigh-speed and reliable service transmission. Especially with the coming of big data age,massive, multi-granularity and heterogeneous service data will swarm into opticalswitching network. At this time, the failure of any link in optical switching network willbring larger amount of data loss, resulting in the communication breakdown extensivelyor for a long time. Obviously, highly survivable performance and dynamicreconfigurable ability are absolutely vital for the construction and application ofbroadband network. The survivability of network directly determines the performanceof network transmission, so the survivability technology is a fundamental andcutting-edge subject about the modern computer and communication network. If theresearch obtains a breakthrough, the related industry will be promoted greatly.Therefore, towards the novel optical switching networks based on self-awaretechnology, around the whole course of service recovery after fault, this thesis mainlycovers the following topics’ research: rapid single fault localization, accurate multi-faultlocalization, protection against key link failures and differentiated protection with trafficaware classification.(1) In view of the existing issues on long fault localization time and high servicedependence, a novel single fault localization mechanism is proposed based on lightpathstatus aware using cluster allocation. The characters of long distance, high speed andtransparency about optical switching network are made comprehensive consideration.According to the constraints of network clustering, two-layer network model isestablished through the minimum dominating set theory. In addition, a new operationcalled “matrix and” is defined in the proposed mechanism. When a link failure occurs,the cluster head and sink node will achieve fast and accurate fault localization via theoperation of “matrix and”. The simulation shows that the fault localization rate andfault localization time are significantly improved with lower complexity and resourcecost.(2) To solve the issue of tracking down multi-link failures fast and accurately forbackbone network, a multi-fault aware localization mechanism based on knight’s tour theory is proposed. First of all, the network model is established according to knight’stour theory. By using the method of the network element abstracting with constraints,the network topology is tried to map into a suitable chessboard with the holes. Then, theprobing signals are sent periodically so that the available light-paths are collecteddynamically at the destination. Finally, the failed links could be determined byiteratively matching the links. The theoretical analysis and simulation experimentdemonstrate that the proposed protocol can effectively localize multi-link failures thatoccur simultaneously with low time complexity in the different network topologies.(3) In order to provide cost-efficient and rapid protection against the dynamical keylink failures, an intelligent p-cycle protection strategy with key link aware based onnetwork coding is proposed in chapter4. Data units are combined from different linksusing network coding method at the on-cycle nodes, and then they are transmitteddownstream for recovering data units lost due to failures. Under static traffic, an integerlinear program (ILP) is formulated to provision the optimal p-cycles. Furthermore,according to the dynamic variation of the link importance degree, a heuristic cycleconstruction algorithm for generating, extending and contracting p-cycle is introducedto achieve intelligent and self-adaptive protection. The key of the proposed protectionstrategy is how to set the key link as a straddling link of the p-cycle as possible. Theexperiments demonstrate that the proposed strategy can guarantee instantaneousrecovery of data units upon the failure of a key link with a low blocking rate andresource cost.(4) To classify IP traffic accurately and provide a reasonable differentiatedprotection according to traffic priority, a network protection strategy with traffic awareclassification is proposed in chapter5. It defines information significance to evaluatethe features of traffic. Combined with Na ve Bayesian theory, a traffic priority model isestablished. Meanwhile, a targeted protection against network failure is providedaccording to the QoS parameters of different priority traffic. The simulation resultsshow that the proposed strategy can achieve accurate traffic classification, and gethigher resource utilization rate and lower blocking rate than traditional differentiatedprotection.