Design And Implementation Of Multiple SDN Controllers Via Zookeeper

Currently, the network has become an important infrastructure of modern social development and technological progress, it is profoundly changing the way of people’s life. However, with the increasing size of the network and network applications, the traditional network architectures is unable to meet the needs of today’s enterprises, operators and users. Especially in the era of cloud computing, multi-tenant, virtual machine migration and other services in data center have increasingly demanding on the network virtualization.SDN(Software Defined Network) technology divide the network into data plane and control plane, data plane is responsible for data forwarding, and the control plane is responsible for developing appropriate forwarding strategies. This architecture which separate controlling and forwarding allows the control layer to get rid of dependence on network devices, providing a flexible and easy programmability. SDN using centralized control which makes it easy to implement resource pooling, dynamic scheduling and expanding. And its open API, the south and north interface, can birth the industrial chain to promote the rapid development of the whole industry.In SDN architecture, controller is extremely important, centralized control also means that if the control plane’s scalability or reliability can’t meet the requirements, the whole network service capabilities will be demoted and even become paralysis.In recent years SDN technology is in transition from the "concept" to "landing", multicontroller design has become one of the core issues of its industrial deployment.This paper has studied the problems in multi-controller system design process and many design principles should weigh,besides some existing programs and theoretical models are also be discussed. This paper proposes a method of using Zookeeper clustered multi-controller architecture management, use Ryu as the controller, use OpenVswitch as OpenFlow switches. In the proposed architecture, Zookeeper service manage the inter-controller status, network-wide view information is also maintained by the Zookeeper server. Each switch is connected to a plurality of controllers, in case of the master controller fails, the other can take over control switches, thus avoiding single points of failure caused by a single controller. While compared to the earlier single controller OpenFlow network architecture, multiple controllers can work together to reduce the load on each controller effectively. The controller can be added dynamically, the entire network topology information is managed by Zookeeper. The proposed architecture using distributed lock to achieve mutually exclusive voting algorithm,and the algorithm can achieve the controller failure handling mechanism. Finally, this paper verified the proposed multi-controller architecture implemented in this article and analyzed herein infrastructure performance bottlenecks, the timing of the elections and the timing of getting the whole network topology information.