首页|面向数据中心的控制器排队性能评估模型研究

面向数据中心的控制器排队性能评估模型研究

扫码查看
软件定义网络SDN(Software-Defined Networking)是通过逻辑控制器对整个网络进行集中控制管理,这种集中式的软件管理保证了网络的开放性且有利于管理,但极易造成网络性能瓶颈.为此,在SDN实际部署网络场景之前,对SDN控制器所在的网络进行性能评估是当前所面临的问题.设计了一种基于排队论在SDN控制器所控制的网络数据中心的性能评估模型,根据SDN在数据中心的Fat-Tree部署模式,及多个OpenFlow交换机发送的Packet-in消息过程特点,分析了Packet-in消息在控制器中的处理时间分布情况.以此为基础,构建了该排队解析模型的理论推导过程和主要性能指标参数,并评估了SDN控制器的处理性能.最后,通过该模型测试了控制器性能,并与现有测试模型进行比较.实验结果表明:该模型比现有测试模型更接近于工具Cbench所测性能指标参数,对实际部署控制器应用网络数据中心具有一定的性能参考价值.
A queueing-based performance evaluation model of SDN controllers in data centers
Software-Defined Networking(SDN)controls and manages the whole network through logic control devices,which keeps the whole network open and manageable on the way of integrated software management.However,it also brings about performance bottlenecks for SDN controllers.Therefore,it is strategic significance to evaluate the performance characteristics for SDN controllers before deploying SDN networks in real situation.In light of this,this paper proposed a performance model of SDN controllers based on queueing theory and analyzed the processing time distribution of the Packet-in messages in SDN controllersaccording to the Fat-Tree topology model of SDN in data centers and the trait that Packet-in messages transmitted from multiple OpenFlow switches.By this way,it constructed the theoretical derivation process and major performance parameters for the queueing model so as to evaluate the processing performance of SDN controller.At last,it tested controller performance with this model and compared it with the existing testing model.Experimental results show that the performance parameters gained by this model are closer to those gotten by Chench,which can provide references for the deployment of SDN controller in data centers.

SDN controllerssoftware-defined networkingperformance evaluationdata centersqueueing models

全秀祥、曹兰

展开 >

闽南师范大学 计算机学院

漳州职业技术学院 电子信息学院,福建 漳州 363000

数据中心 软件定义网络 性能评估 SDN控制器 排队模型

国家自然科学基金项目

62001199

2024

湖北师范大学学报(自然科学版)
湖北师范学院

湖北师范大学学报(自然科学版)

影响因子:0.376
ISSN:2096-3149
年,卷(期):2024.44(1)
  • 17