Research On Key Technologies And Related Problems In Software-Defined Networks

With the development and wide spread of the Internet, traditional net-works are becoming more and more complex and hard to manage, which great-ly hinders innovation and evolution of the networking infrastructure. Software-Defined Networking (SDN) is an emerging networking paradigm that gives hope to change the limitations of current network. By separating the control plane from the data plane, promoting logical centralization of network control and introducing the ability to program the network, SDN simplifies network management and facilitates network innovation, and has recently gained a lot of attention from both academia and industries.However, as an emerging technology, SDN still has a number of problems needed to be solved. Separating the control and forwarding planes is key to the success of SDN, yet it also introduces scalability concerns with respect to the logically centralized control of the network. It is, therefore, necessary to ex-plore the controller architecture so as to improve the scalability and availability of SDN control plane. Distributed controllers have been proposed to address the scalability issue, but how to deploy these controllers in the network is still an open problem. The SDN forwarding devices are entirely controlled by the control plane. In case of failures, connections between switches and controllers could be interrupted, which affects the operation of SDN network. In this sit-uation, mechanisms are needed such that resilient control traffic transmission can be guaranteed. Existing networks consist of a large number of traditional devices, and how to handle the coexistence of SDN and traditional networks, or how to improve the network performance by incrementally deploy SDN nodes in traditional networks is also a challenging task. Based on the above problems, this dissertation studies several key technologies and related problem in SDN, mainly from the aspect of scalability, reliability/resilience and deployment. The main contributions include:1. A distributed controller load balancing architecture, BalanceFlow, is pro-posed to solve the controller load imbalance problem caused by the static controller-to-switch mapping. BalanceFlow enables the super controller to dynamically tune the controller-to-switch mapping according to the load information of all the other normal controllers, and realize controller load balance through switch migration. To overcome the shortages of the current switch migration mechanism, a BalanceFlow-based enhanced switch migration mechanism is proposed, and a controller load balancing algorithm is provided accordingly. Simulation results show that the pro-posed approach can realize tradeoff between load balancing and numbers of switch migrations, and can mitigate the impact of switch migration on controller-to-switch propagation latency.2. A reliability-aware controller deployment strategy is proposed, so as to maximize the reliability of SDN control networks. To characterize the reliability of SDN control networks, a novel metric, called expected per-centage of control path loss is proposed. We then formulate the reliability-aware control deployment problem, prove its NP-hardness, and examine several deployment algorithms. Simulation results show that through s-trategic controller deployment, the reliability of SDN control networks can be significantly improved without introducing unacceptable switch-to-controller latency.3. To achieve resilient control traffic forwarding, this dissertation further investigates the protection of control traffic in SDNs with multiple con-trollers. A control traffic protection scheme that combines both local rerouting and constrained reverse path forwarding protections is proposed. The goal is then to find a set of primary routes for control traffic, called protected control network, where as much control traffic as possible can benefit from the proposed protection scheme. We formulate the protected control network problem and develop an algorithm to solve it. Simula-tion resultsshow that the proposed approach significantly improves the resilience of control traffic.4. Based on the characteristics of hybrid SDN network, where SDN nodes and traditional nodes coexist, an approach that leverages the centralized control ability of SDN is proposed, so as to maximize the traffic flow in hybrid SDN network. We formulate the optimization problem and devel-op a fast Fully Polynomial Time Approximation Scheme (FPTAS) for it. Simulation results show that hybrid SDN networks outperform tradition-al networks, and one can reap a great benefit even with an incrementally deployed SDN.