Failover in Controller Cluster for Software-Defined Networking

• Main Authors: Da-Wei Jiang, 江大衞
• Other Authors: Jiann-Liang Chen
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/n8f9t6

碩士 === 國立臺灣科技大學 === 電機工程系 === 104 === Our daily life has been closely linked with the Internet that it can not be separated. Many applications are combined with network services, resulting in a large number of various kinds of information explosion. Nevertheless, the network architecture of nowadays face many problems. Software-Defined Networking (SDN) technology is the solution. It presents a new network architecture which separates network control plane and data plane. It makes network managers control network easily, and provides a new solution for network applications and future Internet technology. In the SDN network, SDN controller is like a brain, so a positive correlation is inferred between the overall effectiveness of the SDN network and SDN controller. With the scale of networks has been amplified, many researchers consider scalability and single point of failure by a single centralized control in SDN network caused the problems. In order to solve these problems, this work proposes a failover mechanism base on controller cluster in software-defined networking. By using the architecture of controller cluster to improve the scalability and realizing the failover to avoid network interruption from single points of failure (SPOF), the operators can reduce the downtime from network. Thus, controller cluster achieves the goal of making SDN network more stable. This work develops a control plane which combines multiple controllers to form a controller cluster. There are two roles of these controllers: Leader and Member, in this controller cluster. It establishes monitoring mechanisms to monitor the system information of each controller in the cluster. At the same time, it builds failure detection and failover mechanism, in accordance with the timestamp to detect the failed controller, to assign the task to the appropriate controller and to execute the corresponding action. It reduces the entire SDN network downtime, improves the stability of network, makes the data plane to achieve the best processing performance and overcomes the bottleneck of centralized controller in SDN. This work analyzes the failover mechanism base on controller cluster in software-defined networking. First of all, about the failure detection mechanism by the timestamp, the results demonstrate that the proposed mechanism can effectively detect the failed controller. Secondly, the proposed failover mechanism is analyzed according to the results. It shows that the stability of the networks can be enhanced effectively through this mechanism. Moreover, it improves the failover by 22.46% over the achieved using multiple controllers without this mechanism.