基于软件定义网络的网络资源配置研究

发布时间：2019-01-27 22:49

【摘要】：软件定义网络(Software Defined Network, SDN)是近年来提出的一种新型的网络架构形式,而网络功能虚拟化(Network Function Virtualization, NFV)也是最近提出的用虚拟化的方式来实现复杂的网络功能的新概念。使用SDN实现NFV可以达到网络功能完全自动化部署的目标,在当前的网络环境中无疑是有进步意义的。然而当前网络功能极其复杂,使用SDN实现NFV时必然会造成规则的细粒度化,而细粒度规则的处理又会给NFV的硬件承载实体提出较高的要求,本文的研究目标就是在SDN中优化配置网络资源,尤其是CPU等处理机资源,达到全网络资源均衡共享的效果。具体来说,本文的工作主要分为如下三个方面：1.将这类处理资源分配问题用线性规划方式进行建模。该模型的目标函数是使整个网络中分配到的处理任务最多的主机剩余的处理资源最大化,即负载均衡目标,同时又兼顾路由效率；约束条件主要有三类：带宽约束、流量平衡约束和处理资源限制约束。利用线性优化工具,我们求解和分析了特定网络拓扑下的最优解结果。2.我们设计了不拆分网络流的近似算法,对应于对称网络拓扑(Fat-Tree)和不对称网络拓扑(用GT_ITM工具模拟)的场景,分别为：沿最短路径按比例参数分配网络流算法与寻找最优简单路径分配网络流算法。从两个方面(主机分配处理资源的负载均衡度和最大可注入流)对算法效率进行分析,并与线性规划模型的结果进行了比较。3.我们使用OpenFlow协议的多级流表形式设计了对应于我们问题的流表规则,并讨论其如何应对不断变化的网络环境。最后我们对全文进行了总结并提出了进一步的研究方向。
[Abstract]:Software defined network (Software Defined Network, SDN) is a new type of network architecture proposed in recent years, and network function virtualization (Network Function Virtualization, NFV) is a new concept to realize complex network function by virtualization. Using SDN to realize NFV can achieve the goal of full automatic deployment of network functions, which is undoubtedly of great significance in the current network environment. However, the current network functions are extremely complex, and the use of SDN to implement NFV will inevitably lead to the fine-grained rules, and the processing of fine-grained rules will put forward higher requirements for the hardware carrying entities of NFV. The research goal of this paper is to optimize the allocation of network resources in SDN, especially the processor resources such as CPU, so as to achieve the effect of balanced sharing of all network resources. Specifically, the work of this paper is divided into the following three aspects: 1. This kind of processing resource allocation problem is modeled by linear programming. The objective function of the model is to maximize the remaining processing resources of the host which has the most processing tasks assigned to the whole network, that is, the load balancing target, and at the same time to take into account the routing efficiency. There are three kinds of constraints: bandwidth constraints, traffic balance constraints and processing resource constraints. Using linear optimization tools, we solve and analyze the results of optimal solutions in a given network topology. 2. 2. We design an approximate algorithm for unsplit network flows, which corresponds to the scenarios of symmetric network topology (Fat-Tree) and asymmetric network topology (simulated by GT_ITM tools). The network flow allocation algorithm along the shortest path according to the proportional parameters and the algorithm to find the optimal simple path assignment network flow are presented respectively. The efficiency of the algorithm is analyzed from two aspects (load balancing degree of processing resource allocation and maximum injectable flow), and compared with the results of linear programming model. We use the multilevel flow table of OpenFlow protocol to design the flow table rules corresponding to our problem and discuss how to deal with the changing network environment. Finally, we summarize the full text and put forward the further research direction.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.02

【相似文献】