Research On Controller Cluster Of Software Defined Network

With the rapid development of technologies such as virtualization and cloud computing, as well as the constantly emerging of the innovative applications and new network business in fields of social networks and mobile Internet, the demand of Internet for its scale, function and performance is also increasing. The existing protocol architecture of the network device is huge, and the complexity of the management and deployment is high. Therefore, the traditional Internet with the core technology of IPv4 is facing more and more serious technical challenges. The Software Defined Network(SDN) presents a new network architecture in which the data plane and the control plane of the device are decoupling, which provides a new solution for the study of the future network.The research to the controller of the network architecture of SDN has long been the hot topic in the industries, and there have been a lot of solutions to the controller. However, the single controller is facing the performance problem brought by the single point of failure. To solve this problem, it has been started to study the scalability of controller. In this thesis, an analysis is given on the advantages and disadvantages of several current solutions to the controller, and the existing server cluster technology and the virtual cluster technology are also analyzed with a combination of characteristics of SDN. On this basis, this thesis chooses the controller in the open-source project Open Daylight as the object of this study.In this thesis, the features of high availability cluster in the controller of OpenDaylight are firstly studied, in which the features of the controller cluster are studied and analyzed in detail based on a brief analysis on the framework of the controller, including the key features such as cluster communication, data cache, cluster access and information synchronization, etc. At last, a simulation of the underlying network is carried out by means of Mininet, which verifies the consistency function of the cluster feature in the controller of OpenDaylight and tests the reliability of it. On this basis, in order to test the performance and synchronization efficiency of the controller cluster more accurately and provide data support for the actual network deployment of SDN, this thesis designs and realizes a self-learning Mac module with the cluster function in the controller of OpenDaylight. Meanwhile, in the specific scenes, the performance and function of the controller installed such module has been tested, and the result of the test shows that the module can effectively test the processing capacity of the controller itself, and the synchronization function of the module does not affect the whole processing capacity of the controller. At the same time, in a complex network topology, it can be reflected more in the advantages of synchronization of the module, which shows that the establishment of the cluster is more suitable for the management of the complex network topology.