Study On Key Technologies Of IP And Optical Transport Network Uniform Control Based On OpenFlow

With the development and popularization of internet, the rapidly developing IP traffic demands show different characteristics with traditional teleservice, and present new requirements to optical transport network. However, IP layer and optical layer are controlled independently today. There is no thorough coordination mechanism between the two layers, the only interaction is requirement or response. This leads to low transmission quality, high OPEX (Operating Expense) and CAPEX(Capital Expenditure). Thus, flat network and multi-layer uniform controlling are important requirements for transport network. Multi-layer cooperating controlling and some relevant problems have become the research hotspot for both academia and industry in optical transport field.This dissertation studies on the key techniques, including the uniform controlled optical transport network architecture, resource scheduling algorithm, network restoration algorithm and network resource planning algorithm. Some innovative research results have been achieved. The main innovative contribution and work of this dissertation are listed as follows:(1) Multi-layer uniform controlling has become one of the most important requirements for the next generation optical transport network,SDN (Software Defined Networking) and OpenFlow are the key implementation technique. However, there are some problems in realizing a SDN/OpenFlow control plane,such as signal peer invalid problem and Multi-domain network control plane implement problem. In order to solve these problems, this dissertation proposes a signal peer invalid avoiding multi-domain SDN/OpenFlow control plane based on SDN and OpenFlow uniform controlled IP/optical transport network. Firstly, by comparing the centralized control based on SDN/OpenFlow and distributed control based on ASON (Automatic Switch Optical Network)/GMPLS (Generalized Multiprotocol Label Switching), we expound the advantage of SDN/OpenFlow and why centralized control reduces the complexity of control plane and the OPEX. Secondly, we propose the hierarchical control plane which supervises the switches between the host and the backup controller. The stability of control plane is enhanced by controller backup mechanisms and multi-domain network controlling is realized by operation between the inter-domain controller and intra-domain controllers. Lastly, we designed OpenFlow optical switch architecture and optical layer extension of OpenFlow protocol, in order to control optical devices.(2) "Elephant flows" are generated by data synchronization between data centers. Elephant flows require a lot of bandwidth and last longe time. They often lead to link blocking or packet losing. In traditional control mechanism, there is no thorough cooperation mechanism to schedule bandwidth resource between IP and optical layers. In optical transport network, light paths are set up according to IP routing. As a result many add-drop operations have to been done, and network congestion level is enhanced. Scince elephant flows require much bandwidth, there should be a mechanism to guarantee bandwidth fairness among elephant flows while the bandwidth resource is not enough for all the flows. However, elephant flows benefit from as much bandwidth as they can get, but can tolerate transmission delay. This makes it feasible to allocate flexible bandwidth to each flow periodically. According to above problems and the characteristics of elephant flows, we propose an globale routing and bandwidth scheduling algorithm based on fair share of bandwidth and multi-path transmission, in SDN/OpenFlow based uniform controlled IP over OTN (Optical Transport Network) transport network.The proposed algorithm schedules inter-datacenter multi-layer paths and bandwidth periodically, according to the weight and bandwidth requirement of elephant flows. On the one hand, by the cooperation of IP layer and optical layer, elephant flows can be converged and carried by bypassing ODU (Optical channel Data Unit) channels, in order to reduce the impact on other flows (mouse flows); on the other hand, bandwidth resource is allocated according to the weight of elephant flows, in order to increase network throughput, resource utilization and assure fairness.Simulation results show that, compared to high-priority-first mechanism,the method proposed improves average bandwidth allocation ratio by about 10% and also increase the fairness signally. Compared to pure IP layer controlled network, network cost can be reduced by up to 40% with the proposed algorithm in unified controlled IP over OTN network.(3) Since IP flows are carried by optical transport network, a single fault in optical layer often leads to multiple flows break off. In traditional control mechanism, there is no thorough cooperation mechanism to deal with the multi-layer network survivability problem between IP and optical layers. The survivability mechanism is either optical protection which uses redundant resource for backup or IP fast rerouting which calculates restoration path only by IP layer. Both the two methods may lead to low Survivability because of wasting of resources. Furthermore, a mass of restoration requests compete the resource whenever a failure happens. According to above problems, we propose a bulk restoration scheme based on the cooperation between IP and optical layers, in SDN/OpenFlow based uniform controlled IP over OTN transport network.Upon an optical fault, the centralized controller collects the blocked IP flows, calculats restoration paths and allocates bandwidth by considering all the recovering requests. Furthermore, we propose a resource competition avoiding bulk restoration algorithm based on dynamic ordering control. The proposed algorithm controls the ordering to serve the recovering requests according to the bandwidth resource occupation,in order to enhance the survivability of network while keeping the traffic load uniform distribution which avoids resource competition. Simulations show that the results of the proposed algorithm are similar to the optimal solution obtained by MILP (Mixed Integer Linear Programming).Moreover, for the NSFNET reference network, the proposed scheme can signifcantly improve both flow recovery ratio and bandwidth recovery ratio by about 15%.(4) In traditional control mechanism, there is no thorough cooperation resource planning mechanism between IP and optical layer.This leads to transport resource including optical transponders and frequency spectrum can not be efficiently used. Furthermore, there is no effective trade off between transponder occupation and spectrum usage.Generally, transponder occupation and spectrum usage are optimized respectively, while the network cost is not optimized. According to above problems, we quantifiably model multi-layer flexible optical network cost and propose a cost efficient global network resource planning scheme in SDN/OpenFlow based uniform controlled IP over flexible optical transport network. The proposed scheme includs distance-adaptive light path establishment principle, extended auxiliary graph and logical routing caculating method and dynamic ordering control based optimization method. Simulation results show that, the proposed heuristic algorithm has near-optimal performance, close to that obtained with MILP formulations for 6-node network, and the computation time is saved by 4 orders of magnitud. The proposed scheme makes a trade off between transport number and spectrum usage, in order to reduce the total network cost. Moreover, the proposed scheme also reduces spectrum fragmentation generated. Total network cost can be reduced by up to 25%by the proposed scheme, compared to the other schemes.