基于流量调度问题的SDN多控制器结构的研究

发布时间：2018-04-05 00:26

本文选题：软件定义网络　切入点：流量调度　出处：《北京交通大学》2017年硕士论文

【摘要】：网络的大规模发展在为人们带来丰富网络体验的同时,也使网络的业务数量和带宽流量急剧增大,流量调度是提升网络效率和控制能力的一项重要技术。传统的流量调度在全局协调、灵活性等方面存在不可避免的问题。SDN(Software DefinedNetworking,软件定义网络)技术将网络分为应用层、控制层和数据层,极大的提高了网络的灵活性和可扩展性。基于SDN的流量调度可以通过合理的调度数据流以提高带宽利用率从而避免产生拥塞,也可以对网络流量进行精细化控制,动态适应网络流量的实时变化。因此,SDN流量调度对提高网络性能具有重要的研究意义。本文基于SDN流量调度,从SDN多控制器角度对其进行了如下研究工作:(1)详细对SDN中的OpenFlow通信协议进行研究,对SDN交换机和多流表结构及转发流程进行了分析。并从负载均衡的角度对SDN流量调度的两种方式进行了探讨,分析得知SDN流量调度的核心在于对多控制器负载均衡的研究。(2)本文在动态负载均衡算法中加入静态负载信息,由此引出了一种分布式多控制器负载均衡架构 DLCA(Distribute Load-Balance Controller Architecture),该架构由顶层负载均衡系统和底层的控制器层系统组成,每层内部又由各种不同功能的模块构成,并对各个模块的具体实现方式进行了研究,使这一架构能够实现多控制器的负载均衡。(3)本文将Raft选举算法与SDN多控制器进行了结合,使其能够运用选举算法从DLCA架构中的多控制器中选举出负载均衡器。(4)论文对DLCA架构进行了负载均衡算法和选举算法的Mininet仿真。仿真结果表明,DLCA架构的负载均衡器系统能够维持控制器层的负载均衡,并且在某个控制器过载或负载均衡器故障时,选举算法可以从控制器中选举出新的负载均衡器以接替它,从而继续维持系统的流量负载均衡。
[Abstract]:The large-scale development of network not only brings rich network experience, but also increases the number of network services and bandwidth traffic. Traffic scheduling is an important technology to improve network efficiency and control ability.Traditional traffic scheduling has some unavoidable problems in global coordination and flexibility. SDNN Software defined Network (Software defined Network) technology divides the network into application layer, control layer and data layer, which greatly improves the flexibility and expansibility of the network.The traffic scheduling based on SDN can improve the bandwidth utilization and avoid congestion by scheduling the data stream reasonably. It can also fine control the network traffic and dynamically adapt to the real-time change of network traffic.Therefore, SDN traffic scheduling is of great significance in improving network performance.In this paper, based on SDN traffic scheduling, the following research work is done from the view of SDN multi-controller: 1) the OpenFlow communication protocol in SDN is studied in detail, and the structure and forwarding flow of SDN switch and multi-stream table are analyzed.From the point of view of load balancing, this paper discusses two ways of SDN traffic scheduling, and finds out that the core of SDN traffic scheduling is the research of multi-controller load balancing.) in this paper, static load information is added to the dynamic load balancing algorithm.Thus, a distributed multi-controller load balancing architecture, DLCA(Distribute Load-Balance Controller Architecture, is introduced. The architecture consists of a top-level load balancing system and a low-level controller layer system, and each layer is composed of various modules with different functions.And the specific implementation of each module is studied, so that the framework can achieve multi-controller load balancing. (3) this paper combines the Raft election algorithm with the SDN multi-controller.So that it can use the election algorithm to select load balancer from the multiple controllers in the DLCA architecture. This paper simulates the load balancing algorithm and the election algorithm of the DLCA architecture by Mininet.The simulation results show that the load balancer system based on DLCA can maintain the load balance of controller layer, and when a controller overload or load balancer failure, the election algorithm can select a new load balancer from the controller to replace it.Thus continue to maintain the system traffic load balance.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】