• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

基于混合制排队模型的SDN控制器性能评估研究

姜腊林,彭霞,熊兵   

  1. (长沙理工大学计算机与通信工程学院,湖南 长沙 410114)
  • 收稿日期:2015-08-10 修回日期:2015-11-27 出版日期:2017-01-25 发布日期:2017-01-25
  • 基金资助:

    国家自然科学基金(61502056,61303043);湖南省自然科学基金(2015JJ3010,13JJ4052);湖南省教育厅资助科研项目(15B009)

Performance evaluation of SDN
controllers based on hybrid queuing model

JIANG Lalin,PENG Xia,XIONG Bing   

  1. (School of Computer & Communication Engineering,Changsha University of Science & Technology,Changsha 410114,China)
  • Received:2015-08-10 Revised:2015-11-27 Online:2017-01-25 Published:2017-01-25

摘要:

软件定义网络SDN将逻辑控制与数据转发相分离,提高了网络的灵活性和可编程能力,成为近年来未来网络领域的研究热点。SDN在实际应用部署时将面临控制器性能瓶颈的挑战,因而有必要理解SDN控制器的性能特性。为此,首先对SDN控制器中PacketIn消息的到达过程和处理时间进行分析,进而基于排队论提出了一种容量有限的SDN控制器性能评估模型M/M/1/m,推导得出了该模型的性能参数,包括:平均等待队长、平均等待时间、平均队列长度和平均逗留时间。最后,采用控制器性能测试工具Cbench对该模型进行了实验评估。实验结果表明:相对于现有其它模型,该模型的估计时延更接近于实际测量时延,可更精确地描述SDN控制器的性能。

 

关键词: 软件定义网络, 控制器性能评估, 排队模型, 平均逗留时间

Abstract:

oftware defined networking (SDN) expands the flexibility and programmability of networks by decoupling control logic from data forwarding, and becomes a hot topic in the area of future networks. SDN can face the challenge of controller performance bottleneck in its practical deployments, so it is necessary to understand the performance characteristics of SDN controllers. For this end, we first analyze the arrival process and the processing time of PacketIn messages in SDN controllers.Then, we propose a capacitylimited performance evaluation model of SDN controllers M/M/1/m based on the queuing theory. As a further step, we derive its performance parameters, including the average waiting queue length, the average queue time, the average queue length and the average sojourn time. Finally, the proposed performance model is evaluated via experiments by using the controller performance measurement tool Cbench. Experimental results indicate that the estimated delays of our model are closer to the measured delays than those of other existing models, thus providing a more accurate description of SDN controller performance.

Key words: software defined networking (SDN), controller performance evaluation, queuing model, average sojourn time