Research On Several Key Technologies Of Gmpls-Based Intelligent Multi-Layer Multi-Domain Optical Networks

With the fast development of Internet and the rapid growth of broadband data service, the optical transport network is experiencing a revolution towards a new network with larger scale, higher reliability and more intelligence. GMPLS technology has enabled the multi-granularity switching and multi-domain/multi-layer architecture has strengthened the network scalability. In the network revolution, it is necessary to cooperate and integrate several technologies to have the optimal mechanism of resource allocation and flexible scheduling, fast, reliable and precise route computation and path establishment signaling, and robust survivability for the network control and management. Therefore, it is a hot research point that how to utilize the network resource effectively to build high reliable optical network, which will not only play an important practical role, but also be of great significance.Focus on the new requirements and technology difficulty of the multi-layer and multi-domain optical network, the research work of this dissertation is based on the national863project named DREAMSCAPE. This dissertation mainly focuses on the research fields such as routing and resource assignment, network survivability, spectrum assignment and resource defragmentation based on the GMPLS technology. The main work and innovative ideas of this dissertation are summarized as follows:Firstly, as the traditional routing and distributed wavelength assignment (R+DWA) scheme is easy to cause the wavelength resource reservation confliction, a PCE-based dynamic resource backup reservation scheme has been proposed in this dissertation. This scheme makes use of the central calculation superiority of PCE and the advantages of both the first fit and random fit algorithms to reserve wavelength resource for the lightpath. In order to reduce the network resource occupation of the backup wavelength, a dynamic threshold is defined according to the hops of the path and the number of available wavelengths to decide whether to reserve a backup wavelength. Simulation results show that the proposed scheme achieves lower blocking probability. When the traffic load reaches160Erlang and threshold is set to5, the blocking probability is only3.05%.Secondly, considering the different impacts on the service with different bandwidth and continuity in SLICE, a service-classifying-based routing and spectrum assignment scheme is proposed that computes the route by setting different weights for different services with OSPF. Simulation results show that the scheme can allocate the bandwidth resource reasonably and archives lower total blocking probability. Compared with multiplicative weight scheme, additive weight scheme and the scheme only considering hop number, the blocking probability reduces1%-3%under heavy traffic load.Thirdly, a service flow monitoring-based elastic modulation scheme is proposed in order to solve the high blocking probability problem in heavy traffic load. When service flow reaches the certain threshold, the higher order modulation is adopted to compress the service bandwidth on condition that the Quality of Service is guaranteed. Simulation results show that the proposed scheme not only achieves the balance between QoS and blocking probability but also improves the network performance obviously by setting a proper threshold.Fourthly, a spectrum availability-based resource defragmentation scheme is proposed in allusion to that path is established and teared down frequently that leads to lower resource utilization rate and whether the spectrum is moved or not depends on the link availability. Simulation results show that compared with the unallocated scheme the blocking probability of the allocated scheme reduces1.7%under the circumstance of heavy traffic load in the whole network.Fifthly, considering the difficulty of path computation and establishment in a large scale network, a Dual Routing Engine architecture combining ASON, GMPLS and PCE is proposed. In the DREAM architecture, two routing algorithms DRE-FPC and HDRE-BRPC are proposed and tested, both of which have combined the advantages of centralization and distribution. The author also proposes the scheme to solve the distributed resource assignment confliction problem in large scale network and constructs the experimental platform, on which the architecture, mechanisms and algorithms aformentioned are tested.The test results show that the two algorithms perform better than traditional way in the aspect of resource utilization rate, establishment delay and blocking probability. The results show that the average establishment delay of two routing schemes reduces about437-538ms and the blocking probability reduces about1%-3%comparing with HR scheme.In addition, a survivability evaluation mechanism in multi-domain optical network is proposed which extends relevant protocols of control plane to measure different restoration time in order to describe the network survivability performance. Based on the achievement above, an international standard draft has been proposed and updated to the second version, of which the life time is more than one year.