SDN控制平台方案研究与实现

发布时间：2019-06-27 12:08

【摘要】：随着云计算、物联网、移动互联网、社交网络等新技术、新模式、新应用的快速发展,.互联网规模也急剧扩大,数据量快速增长。而传统网络的弊端在处理以大数据、智能分析、网络虚拟化为代表的新业务、新应用时日益显现。臃肿复杂的网络结构、不断攀升的成本以及僵化的网络管理使得网络研究者开始探寻新的网络架构设计。在此背景下,软件定义网络应运而生。软件定义网络(SDN)是一种新兴的网络架构,它将网络控制从转发中分离出来,从而可以通过软件编程实现对数据转发规则的控制。在SDN中,控制平台同时简化了节点设备(节点设备只负责转发,转发行为由控制平台控制)和网络控制逻辑(只需调用控制平台的编程接口,而不用处理底层网络状态),并且还可以支持广泛的功能,因此控制平台在SDN架构中是一个十分关键的组成部分。只有设计一个高效可扩展的控制平台才能保证SDN网络的良好性能,而这就是本文的主要研究目标。本文首先介绍了软件定义网络的架构和核心技术,分析比较了现有的SDN控制平台相关研究,明确了研究重点。之后本文提出了一个SDN控制平台方案。该方案采用分布式架构,实现了SDN控制器间接口,保证了多个控制器间状态的一致性；只对重要的流进行处理,减轻了控制平台和转发设备的负荷；主动下发预先设置好的策略,降低了控制开销和流初始建立时延。之后本文将SDN控制平台应用到数据中心场景中,利用控制平台控制管理数据中心中的大象流。本文将这个数据中心中的SDN控制平台具体应用称为数据中心大象流控制管理系统。该系统只对大象流进行处理,对于老鼠流使用最基本的ECMP,因此系统可以获得很好的可扩展性。同时本文提出了一个加权多路径路由算法,其按照当前网络负载情况动态计算各个路径的权重。然后系统将检测到的大象流,按照该算法计算的权重比率,分割到多个路径上以平衡负载。本文实现了一个系统原型,并在OpenFlow测试网络中验证了本文提出的机制的有效性。最后本文对SDN控制平台未来发展进行了展望。
[Abstract]:With the rapid development of cloud computing, Internet of things, mobile Internet, social network and other new technologies, new models, new applications. The scale of the Internet has also expanded rapidly, and the amount of data has grown rapidly. However, the disadvantages of the traditional network are increasingly emerging in dealing with the new business represented by big data, intelligent analysis and network virtualization, and the new application is becoming more and more obvious when dealing with the new business represented by big data, intelligent analysis and network virtualization. Bloated and complex network structure, rising cost and rigid network management make network researchers begin to explore new network architecture design. In this context, the software definition network emerges as the times require. Software defined network (SDN) is a new network architecture, which separates network control from forwarding, so that the control of data forwarding rules can be realized by software programming. In SDN, the control platform simplifies the node device (the node device is only responsible for forwarding, the forwarding behavior is controlled by the control platform) and the network control logic (only needs to call the programming interface of the control platform instead of dealing with the underlying network state), and can also support a wide range of functions, so the control platform is a very key part of the SDN architecture. Only by designing an efficient and extensible control platform can the good performance of SDN network be guaranteed, which is the main research goal of this paper. This paper first introduces the architecture and core technology of software definition network, analyzes and compares the existing research of SDN control platform, and clarifies the research focus. Then a scheme of SDN control platform is proposed in this paper. The scheme adopts distributed architecture to realize the interface between SDN controllers, which ensures the consistency of states among multiple controllers; only the important flow is processed, which reduces the load of the control platform and forwarding equipment; and the pre-set strategy of actively sending reduces the control overhead and the initial establishment delay of the flow. Then this paper applies the SDN control platform to the data center scenario, and uses the control platform to control and manage the elephant flow in the data center. In this paper, the application of SDN control platform in this data center is called Elephant flow Control Management system. The system only deals with elephant flow, and uses the most basic ECMP, for mouse flow, so the system can achieve good scalability. At the same time, a weighted multi-path routing algorithm is proposed, which calculates the weight of each path dynamically according to the current network load. Then the detected elephant flow is divided into multiple paths to balance the load according to the weight ratio calculated by the algorithm. In this paper, a system prototype is implemented, and the effectiveness of the proposed mechanism is verified in OpenFlow test network. Finally, the future development of SDN control platform is prospected.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06

【共引文献】