基于SDN的智能网络控制机制研究

发布时间：2018-07-01 13:06

本文选题：软件定义网络 + 多控制器　；参考：《北京邮电大学》2017年硕士论文

【摘要】：随着互联网的不断发展,人们对于网络的性能需求越来越高。现在的TCP/IP网络面临着复杂的网络管理、封闭的网络设备等问题。软件定义网络(Software-defined Network, SDN)是一种新型网络创新架构,可以有效地解决当前网络中难以解决的问题。随着SDN网络规模地逐渐扩大,单个逻辑集中的控制器会遇到单点故障以及可扩展性问题。而分布式控制系统是解决可扩展性问题的有效方案。分布式控制系统中设计到多个控制器,在大型网络中如何部署控制器,也是个需要解决的挑战。多控制器部署问题中,交换机依赖于控制器来进行数据转发处理,因此交换机和控制器之间的链路稳定是一个很重要的因素。因此本论文目标是找到一种控制器的部署方案,保证交换机和控制器之间通信的可靠性,同时尽可能地减小交换机和控制器之间的通信延时。同时,在中小型网络中部署多控制器时,除了传播时延,控制器的响应时间也是影响QoS一个很重要的因素。因此,本论文研究在实际网络部署时,如何使用合理的网络资源来满足QoS要求。本文针对以上两个方面分别进入了深入研究,主要贡献如下:(1)基于对SDN分布式控制器和多控制器部署算法调研的基础上,研究了转发平面故障对于网络性能的影响,尤其是交换机和控制器性之间的可靠性,提出一种基于稳定性的cycle k-medoids算法。(2)本论文在真实拓扑上进行仿真,结果表明本文提出的cycle k-medoids算法明显地提高交换机和控制器通信的可靠性,同时保持比较低的时延。(3)研究了多控制器部署时的服务质量(Quality of Service,QoS)问题。本论文对SDN中流的处理过程进行了建模,分析了控制器对于流的响应时间及其影响因素。同时考虑控制器的负载和传播时延,提出一种保证QoS的启发式算法。(4)本论文在真实拓扑上进行仿真,指出在一定范围下提高控制器容量和时延上限可以减小所需部署的控制器数量。这对于实际网络中部署有一定的指导作用。
[Abstract]:With the continuous development of the Internet, the demand for network performance is becoming higher and higher. The current TCP / IP network is faced with complex network management, closed network equipment and other problems. Software-defined Network (SDN) is a new network innovation architecture, which can effectively solve the difficult problems in the current network. As the scale of SDN network expands gradually, the controller in a single logic set will encounter single point failure and scalability problems. Distributed control system is an effective solution to scalability problem. It is also a challenge to design multiple controllers in distributed control systems and how to deploy controllers in large networks. In the problem of multi-controller deployment, the switch relies on the controller for data forwarding, so the link stability between the switch and the controller is an important factor. Therefore, the purpose of this thesis is to find a controller deployment scheme to ensure the reliability of the communication between the switch and the controller, and to reduce the communication delay between the switch and the controller as much as possible. In addition to propagating the delay, the response time of the controller is also an important factor affecting QoS when deploying multiple controllers in small and medium-sized networks. Therefore, this paper studies how to use reasonable network resources to meet QoS requirements in actual network deployment. The main contributions of this paper are as follows: (1) based on the investigation of SDN distributed controller and multi-controller deployment algorithm, the effect of forwarding plane fault on network performance is studied. Especially the reliability between switch and controller, a stability based cycle k-medoids algorithm is proposed. (2) this paper simulates the real topology. The results show that the proposed cycle k-medoids algorithm can obviously improve the reliability of the communication between the switch and the controller, while maintaining a relatively low delay. (3) the quality of Service (QoS) problem in multi-controller deployment is studied. In this paper, the process of flow processing in SDN is modeled, and the response time of controller to flow and its influencing factors are analyzed. At the same time, considering the load and propagation delay of the controller, a heuristic algorithm is proposed to guarantee QoS. (4) this paper simulates in real topology. It is pointed out that increasing the upper limit of controller capacity and delay can reduce the number of controllers to be deployed in a certain range. This has a certain guiding role for the actual network deployment.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】