SDN集群控制器高可用性研究与设计
发布时间：2019-07-10 09:59
【摘要】：软件定义网络(Software Defined Networking,SDN)作为未来网络的主要研究方向,转发与控制分离的网络架构也越来越深入人心,SDN通过集中控制可以实现业务的快速部署,也使得整个网络资源的利用更加合理。集中式控制层面虽然带来了可控性、安全性和合理性等优点,但随着业务规模不断的增大及应用场景的复杂化,也产生了很多扩展性、可靠性、高可用性问题。本文深入研究了 SDN控制器的可扩展性问题,创造性得提出了一种控制器可扩展方案,并对该方案下的高可用性相关问题进行研究。首先本文详细设计了一种SDN集群控制器高可用性架构;其次通过对所设计的高可用架构进行研究,并针对各层的特性,提出三种不同的故障检测方法。针对根控制器管理系统,主要采用了主备检测方法;对于根控制器故障的检测,提出了过半检测方法;针对本地控制器的故障检测,提出了循环检测方法和控制器间自适应延时算法。主要实验仿真结果显示所提出方法可以快速检测故障并且提高检测准确度。再者,本文对主控制器选举算法进行深入研究,并提出了一种高可用的主控制器选举算法,实验仿真结果显示所设计的算法可以有效地选举出高可用的主控制器。
文内图片：
图片说明：图１－１所示，Ｈｙｐｅ巧ｌｏｗＷ是一种基于ＮＯＸ的分布式控制设计方案
[Abstract]:As the main research direction of the future network, the network architecture of the separation of forwarding and control is becoming more and more popular. SDN can realize the rapid deployment of services through centralized control, and make the utilization of the whole network resources more reasonable. Although the centralized control level brings the advantages of controllability, security and rationality, with the continuous increase of business scale and the complexity of application scenarios, there are a lot of scalability, reliability and high availability problems. In this paper, the scalability of SDN controller is deeply studied, a controller extensibility scheme is proposed creatively, and the high availability related problems under this scheme are studied. Firstly, a high availability architecture of SDN cluster controller is designed in detail. Secondly, three different fault detection methods are proposed according to the characteristics of each layer by studying the designed high availability architecture. For the root controller management system, the main and standby detection method is mainly adopted; for the root controller fault detection, more than half detection method is proposed; for the local controller fault detection, the cyclic detection method and the adaptive delay algorithm between controllers are proposed. The main experimental simulation results show that the proposed method can detect faults quickly and improve the detection accuracy. Furthermore, the main controller election algorithm is deeply studied in this paper, and a highly available main controller election algorithm is proposed. The experimental simulation results show that the designed algorithm can effectively elect the highly available main controller.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.02
本文编号：2512529
文内图片：
图片说明：图１－１所示，Ｈｙｐｅ巧ｌｏｗＷ是一种基于ＮＯＸ的分布式控制设计方案
[Abstract]:As the main research direction of the future network, the network architecture of the separation of forwarding and control is becoming more and more popular. SDN can realize the rapid deployment of services through centralized control, and make the utilization of the whole network resources more reasonable. Although the centralized control level brings the advantages of controllability, security and rationality, with the continuous increase of business scale and the complexity of application scenarios, there are a lot of scalability, reliability and high availability problems. In this paper, the scalability of SDN controller is deeply studied, a controller extensibility scheme is proposed creatively, and the high availability related problems under this scheme are studied. Firstly, a high availability architecture of SDN cluster controller is designed in detail. Secondly, three different fault detection methods are proposed according to the characteristics of each layer by studying the designed high availability architecture. For the root controller management system, the main and standby detection method is mainly adopted; for the root controller fault detection, more than half detection method is proposed; for the local controller fault detection, the cyclic detection method and the adaptive delay algorithm between controllers are proposed. The main experimental simulation results show that the proposed method can detect faults quickly and improve the detection accuracy. Furthermore, the main controller election algorithm is deeply studied in this paper, and a highly available main controller election algorithm is proposed. The experimental simulation results show that the designed algorithm can effectively elect the highly available main controller.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.02
【参考文献】
相关期刊论文 前9条
1 杨柳;马晓琛;;集群技术研究[J];信息通信;2014年02期
2 林萍萍;毕军;胡虹雨;蒋小可;;一种面向SDN域内控制平面可扩展性的机制[J];小型微型计算机系统;2013年09期
3 韩言妮;覃毅芳;慈松;;未来网络虚拟化关键技术研究[J];中兴通讯技术;2011年02期
4 朱良勇;;基于TCP协议的双机热备份系统[J];现代防御技术;2008年04期
5 罗娟,曹阳,郑刚,张俊新;高可用性软件的设计与实现[J];计算机工程;2004年08期
6 姜坚华;双机热备系统的技术研究和具体实现[J];微型电脑应用;2004年03期
7 薛军,李增智,王云岚;负载均衡技术的发展[J];小型微型计算机系统;2003年12期
8 成汝震;尚志恩;刘宏忠;张运凯;;分布式系统数据一致性处理的研究[J];计算机科学;2001年08期
9 吴辉;基于有向生成树的分布式选举算法[J];计算机研究与发展;1995年08期
本文编号：2512529
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2512529.html
