SDN多控制器部署及动态迁移机制研究

发布时间：2018-05-22 20:07

本文选题：软件定义网络 + 控制层　；参考：《西南交通大学》2017年硕士论文

【摘要】：软件定义网络(Software Defined Network,SDN)是一种新型的网络架构。与传统的IP网络相比,该架构将控制层与转发设备进行解耦并将其集中到独立的控制器中。这种改进大大提高了网络的开放性,有利于实现对网络的灵活管理。作为一种新兴架构,SDN仍存在许多尚未解决的技术难题。本文研究了 SDN网络的可扩展性和动态管理性的问题。通过对SDN中数据传输和路径请求的过程进行研究,分析了影响网络扩展性的几个关键因素,并利用仿真对不同控制器部署方案的优缺点进行了对比。为进一步提高网络的性能,本文设计了一种新的控制器和交换机映射关系的局部动态调整方案,并构建模型进行了验证。首先将改进后的Jackson模型应用到SDN网络中,研究了数据包在网络中的传输原理,并且从理论上分析了控制器对数据包的处理时延分布和负载分布情况。利用MATLAB建立了仿真模型,通过不同的请求率和服务率对模型进行了验证。其次,本文分析了多控制器部署中常用的几种算法,例如经典的k-center和k-median,并对这些算法的不足之处进行了改进,设计了新型的去边缘化(DM)算法。DM算法以网络的连通性和稳定性为优化目标进行迭代,在多节点中寻找最优的控制节点组合,并对优化后的部署方案进行了仿真验证。最后,针对网络流的动态变化特性,对静态部署方案进行了改良,研究并设计了基于自适应优化算法的交换机动态迁移策略,使网络能够以较小的代价实现对控制器和交换机之间映射关系的灵活调整,从而适应动态流需求。仿真结果表明,该方案能够实现多控制器的负载均衡,提高控制器资源的利用率。
[Abstract]:Software definition Network Defined (SDN) is a new network architecture. Compared with the traditional IP network, the architecture decouples the control layer from the forwarding device and centralizes it into a separate controller. The improvement greatly improves the openness of the network and facilitates the flexible management of the network. As a new architecture, SDN still has many unsolved technical problems. This paper studies the scalability and dynamic management of SDN networks. By studying the process of data transmission and path request in SDN, several key factors affecting network expansibility are analyzed, and the advantages and disadvantages of different controller deployment schemes are compared by simulation. In order to further improve the performance of the network, this paper designs a new local dynamic adjustment scheme of the mapping relationship between controller and switch, and builds a model to verify it. Firstly, the improved Jackson model is applied to the SDN network, and the transmission principle of the data packet in the network is studied, and the delay distribution and the load distribution of the data packet processing by the controller are analyzed theoretically. The simulation model is established by using MATLAB, and the model is verified by different request rate and service rate. Secondly, this paper analyzes several algorithms commonly used in multi-controller deployment, such as classical k-center and k-median, and improves the shortcomings of these algorithms. A new de-marginalizing DM.DM algorithm is designed to iterate over the connectivity and stability of the network to find the optimal control node combination in multi-nodes. The optimized deployment scheme is verified by simulation. Finally, according to the dynamic characteristics of network flow, the static deployment scheme is improved, and the switch dynamic migration strategy based on adaptive optimization algorithm is studied and designed. The network can adjust the mapping relationship between controller and switch at a lower cost, so as to adapt to the demand of dynamic flow. The simulation results show that the proposed scheme can balance the load of multiple controllers and improve the utilization of controller resources.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.02

【相似文献】