基于OVS的SDN可靠性研究

发布时间：2018-10-10 06:11

【摘要】：SDN (Software Defined Networking,软件定义网络)成为未来网络发展的重要方向之一,核心思想是实现控制与转发分离以及开放的可编程性,并不断优化和重构现有网络。SDN可靠性问题是SDN发展过程中的核心问题之一,其工作目标是在不可避免的网络故障中,尽可能减小网络故障带来的网络性能影响。在现有SDN可靠性问题研究中,存在故障通信空白期长、故障原因多样化、域内域间可靠性和故障恢复信息更新可靠性等多方面挑战。本文在对现有研究工作中存在的不足点和挑战展开深入的分析后,并将目前数据中心等SDN场景中广泛应用的软件交换机OVS (Open vSwitch,软件交换机)作为架构基础,最终形成了基于OVS的SDN可靠性解决方案,本文主要的创新点如下:1)提出基于OVS的SDN域内可靠性机制。本文通过拓展OpenFlow连接状态匹配域,实现数据包在控制链路的不同连接状态下状态匹配功能,从而使得交换机快速感知故障并及时平滑切换至灾备状态;此外,本文在OpenFlow中拓展自学习动作,不仅将一些基本的学习与转发功能在数据层面高效执行,减小控制器负载与延迟,而且保障在网络在控制器不可用场景下的基本通信功能。该方案不仅提高SDN域内网络可靠性,而且创造性的设计符合SDN思想,保留SDN的可拓展性和灵活性。2)提出基于OVS的SDN域间可靠性机制。基于域内设计的状态匹配机制,本文另外设计了转发机制和控制协同等机制。采用这些机制,本文从数据包入域和出域两个维度,巧妙的解决SDN控制器失效场景下的数据包跨域可靠性问题。3)提出基于OVS的故障恢复的信息更新保障机制。本文分别从控制器信息更新和交换机信息更新两个层面,利用设计的信息更新保障机制,解决了故障恢复后等出现的流表、拓扑信息的过期和老化等问题。4)为了证明设计方案的可行性,本文基于OVS对各个机制模块化,并详细描述其实现思路。为了进行性能评估,本研究开展多组实验进行验证分析,实验结果表明,无论是域内还是域间,控制器失效时的故障切换时间保持在100ms级别,丢包率低且具有较为平稳的吞吐量。因此,本文设计方案相比传统方案,可以大大减小控制器不可用等故障带来的网络影响,提高了 SDN的网络可靠性。
[Abstract]:The definition of network by SDN (Software Defined Networking, software has become one of the important directions of network development in the future. The core idea is to realize the separation of control and forwarding and open programmability. And continuously optimizing and reconstructing the existing network. SDN reliability problem is one of the core problems in the development of SDN. Its goal is to minimize the network performance impact caused by the network failure in the inevitable network failure. In the existing research of SDN reliability, there are many challenges such as long gap period of fault communication, diversification of fault causes, inter-domain reliability and reliability of fault recovery information update. After deeply analyzing the shortcomings and challenges in the existing research work, this paper takes the OVS (Open vSwitch, software switch, which is widely used in the current data center and other SDN scenarios, as the foundation of the architecture. Finally, the SDN reliability solution based on OVS is formed. The main innovation of this paper is as follows: 1) A new reliability mechanism in SDN domain based on OVS is proposed. In this paper, we extend the OpenFlow connection state matching domain to realize the state matching function of the data packet in different connection states of the control link, so that the switch can quickly perceive the fault and switch smoothly to the disaster preparedness state in time. In this paper, we extend the self-learning action in OpenFlow, not only carry out some basic learning and forwarding functions efficiently at the data level, reduce the load and delay of the controller, but also guarantee the basic communication function of the network in the scenario where the controller is not available. This scheme not only improves the network reliability in SDN domain, but also creatively designs in line with the SDN idea, keeping the expansibility and flexibility of SDN. 2) A OVS based interdomain reliability mechanism for SDN is proposed. Based on the state matching mechanism designed in the domain, the forwarding mechanism and control cooperation mechanism are also designed in this paper. Using these mechanisms, this paper proposes an information update guarantee mechanism for fault recovery based on OVS from the two dimensions of data packet entry and outfield, and skillfully solves the packet cross-domain reliability problem in the failure scenario of SDN controller. In this paper, from the two levels of controller information update and switch information update, the flow table after fault recovery is solved by using the designed information update safeguard mechanism. In order to prove the feasibility of the design scheme, this paper modulates each mechanism based on OVS, and describes its implementation idea in detail. In order to evaluate the performance of the controller, a number of experiments are carried out to verify and analyze the performance. The experimental results show that the failover time of the controller is maintained at the 100ms level when the controller fails, whether within or between domains. Low packet loss rate and relatively stable throughput. Therefore, compared with the traditional scheme, the proposed scheme can greatly reduce the network impact caused by the non-availability of controller and improve the network reliability of SDN.
【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】