基于SDN的物联网拥塞控制研究

发布时间：2018-01-14 22:24

本文关键词：基于SDN的物联网拥塞控制研究　出处：《湖北工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着物联网(IoT)快速发展,接入网络的设备持续增加,带来业务数据流量爆发式的增涨。而传统IP网络采用分布式控制方法,缺乏全局网络信息,不能很好满足大量不同业务类型需求,使现有网络拥塞控制陷入瓶颈。软件定义网络技术(Software Defined Networking,SDN)通过将数据转发层与控制层解耦,摆脱了传统拥塞控制的固有限制,实现更优化、适应性更强的拥塞控制,进而解决物联网对传输网络的需求。论文对传统网络中存在节点的阻塞问题和网络资源分配不均衡的问题,设计了基于SDN的物联网中数据转发层和控制层拥塞控制方案。其主要工作如下:针对基于SDN的物联网数据转发节点拥塞的问题,分析了目前尾部丢弃、随机早起检测(RED)、公平排队(SFQ)等主流的队列管理算法在队列的公平性和分组优先级等方面的不足,设计了一种基于区分业务流的主动队列管理算法。该算法利用马尔可夫决策预测下一时刻的状态,并用层次分析法计算不同业务流的队列属性权重。然后基于切换动作和队列状态构建期望回报函数,采用逐次逼近的迭代得到长期期望回报最大的队列切换选择。最后通过仿真分析,对提出的方案进行了验证。结果表明,该方法可有效降低数据业务流在转发节点的阻塞率,提高网络资源利用率。针对基于SDN的物联网网络资源分配不均衡以及资源利用率较低的问题,在控制层设计了一种基于最小费用流的虚拟映射算法来实现流量优化控制。通过对SDN控制器的流量转发的特点进行分析,建立虚拟网络映射的数学模型,将最小费用流应用于流量路径的寻优,以实现对网络的集中控制。根据改进单纯形法和k短路法构造一种禁忌搜索算法以解决上述规划问题。最后给出该算法的仿真分析,结果表明,该算法相较于传统的SPF(Shortest Path First)算法分布均衡性更好,资源利用率更高。
[Abstract]:With the rapid development of IoT, the number of devices connected to the network continues to increase, which leads to the explosive increase of service data flow. The traditional IP network uses distributed control method and lacks global network information. It can not meet the needs of a large number of different business types, which makes the existing network congestion control into a bottleneck. Software Defined Networking. By decoupling the data forwarding layer and the control layer, SDN gets rid of the inherent limitation of the traditional congestion control and realizes more optimized and more adaptive congestion control. Then solve the demand of the Internet of things for the transmission network. The traditional network of nodes in the problem of congestion and uneven allocation of network resources. This paper designs the congestion control scheme of data forwarding layer and control layer in Internet of things based on SDN. The main work is as follows: aiming at the congestion problem of data forwarding node in Internet of things based on SDN, this paper analyzes the current tail discard. The main queue management algorithms, such as random early detection (red), fair queuing (SFQ) and so on, are insufficient in queue fairness and packet priority. An active queue management algorithm based on differentiated traffic flow is designed, which uses Markov decision to predict the state of the next moment. The queue attribute weights of different traffic flows are calculated by AHP, and then the expected return function is constructed based on the switching action and queue state. Finally, the proposed scheme is verified by simulation analysis, and the results show that the proposed scheme is feasible. This method can effectively reduce the blocking rate of data traffic in forwarding nodes and improve the utilization of network resources. Aiming at the problem of unbalanced resource allocation and low resource utilization in the Internet of things based on SDN. In the control layer, a virtual mapping algorithm based on minimum cost flow is designed to realize optimal flow control. By analyzing the characteristics of traffic forwarding of SDN controller, the mathematical model of virtual network mapping is established. The minimum cost flow is applied to the optimization of the flow path. In order to realize the centralized control of the network, a Tabu search algorithm is constructed according to the improved simplex method and k-short circuit method to solve the above planning problem. Finally, the simulation analysis of the algorithm is given, and the results show that the algorithm can solve the problem. Compared with the traditional SPF(Shortest Path first algorithm, this algorithm has better distribution balance and higher resource utilization.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5;TP391.44

【相似文献】