作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2022, Vol. 48 ›› Issue (4): 106-112. doi: 10.19678/j.issn.1000-3428.0061217

• 人工智能与模式识别 • 上一篇    下一篇

基于成本的服务链组成与部署联合优化策略

刘雅丽, 史久根   

  1. 合肥工业大学 计算机与信息学院, 合肥 230009
  • 收稿日期:2021-03-22 修回日期:2021-05-07 发布日期:2021-05-11
  • 作者简介:刘雅丽(1996—),女,硕士研究生,主研方向为网络功能虚拟化;史久根,副教授。
  • 基金资助:
    国家重大科学仪器设备开发专项(2013YQ030595)。

Joint Optimization Strategy for Cost-Based Service Chain Composition and Deployment

LIU Yali, SHI Jiugen   

  1. School of Computer and Information, Hefei University of Technology, Hefei 230009, China
  • Received:2021-03-22 Revised:2021-05-07 Published:2021-05-11

摘要: 网络功能虚拟化(NFV)通过将虚拟网络功能(VNF)部署在虚拟设备中,提高了网络管理的灵活性,但随着服务需求的扩大,网络供应商消耗的运营支出(OPEX)也不断增加。由于VNF改变流大小的特性、VNF间的依赖性以及组成和部署方式的复杂性,面向OPEX的VNF组成和部署问题充满挑战。提出一种面向成本的虚拟网络链组成和部署联合优化策略,将节点映射成本、链路映射成本、激活成本和能耗成本公式化为OPEX,构建混合整数非线性规划模型。为分析影响成本的不同因素,同时提高特殊依赖情况下的处理效率,根据不同依赖关系将VNF请求集分为完全无序、部分有序和完全有序VNF集合进行分析,并设计3种相应优化算法。实验结果表明,在完全无序、部分有序情况下,算法性能优于首次适应算法、随机拟合算法等同类算法,对于完全有序算法,当节点资源配比在50%以上时,可获得小规模网络下近似线性规划模型精确解的方案。

关键词: 网络功能虚拟化, 虚拟网络功能链组成, 虚拟网络功能链放置, 运营支出, 服务功能链

Abstract: Network Function Virtualization(NFV) improves network management flexibility by deploying Virtual Network Functions(VNF) in virtual devices.With the continuous expansion of service demand, the Operational Expenditure(OPEX) consumed by network providers is increasing.The efficient chaining and placing of VNF is challenging based on their traffic-changing effects and dependency relationships, as well as the diversity of available server choices.This paper proposes a cost-oriented joint optimization strategy for virtual network chain composition and deployment, formulates node mapping cost, link mapping cost, activation cost, and energy consumption cost as OPEXs, and establishes a mixed-integer nonlinear programming model.To analyze the different factors that affect the cost and improve the processing efficiency of special dependence, the VNF request set is divided into non-ordered, partially ordered, and totally ordered VNF according to different dependencies and three corresponding optimization algorithms are designed.Experimental results demonstrate that the performance of the proposed algorithm is better than that of first time adaptive algorithm and random fitting algorithms in the cases of complete non-order and partial order.When the node resource ratio is greater than 50%, the totally ordered VNF algorithm can obtain a solution that approximates the exact solution of the linear programming model in a small-scale network.

Key words: Network Function Virtualization(NFV), Virtual Network Functions(VNF) chain composition, VNF chain placement, Operational Expenditure (OPEX), Service Function Chaining (SFC)

中图分类号: