通信学报 ›› 2022, Vol. 43 ›› Issue (2): 131-142.doi: 10.11959/j.issn.1000-436x.2022038

• 学术论文 • 上一篇    下一篇

QL-STCT:一种SDN链路故障智能路由收敛方法

李传煌, 陈泱婷, 唐晶晶, 楼佳丽, 谢仁华, 方春涛, 王伟明, 陈超   

  1. 浙江工商大学信息与电子工程学院(萨塞克斯人工智能学院),浙江 杭州 310018
  • 修回日期:2022-01-17 出版日期:2022-02-25 发布日期:2022-02-01
  • 作者简介:李传煌(1980-),男,江西九江人,博士,浙江工商大学教授、硕士生导师,主要研究方向为软件定义网络、开放可编程网络、边缘计算、人工智能应用
    陈泱婷(1998-),女,浙江绍兴人,浙江工商大学硕士生,主要研究方向为软件定义网络、人工智能应用等
    唐晶晶(1998-),女,浙江杭州人,浙江工商大学硕士生,主要研究方向为软件定义网络、人工智能应用等
    楼佳丽(1998-),女,浙江东阳人,浙江工商大学硕士生,主要研究方向为软件定义网络、人工智能应用等
    谢仁华(1997-),男,浙江临海人,浙江工商大学硕士生,主要研究方向为软件定义网络、人工智能应用等
    方春涛(1995-),男,浙江建德人,浙江工商大学硕士生,主要研究方向为软件定义网络、人工智能应用等
    王伟明(1964-),男,浙江遂昌人,博士,浙江工商大学教授、硕士生导师,主要研究方向为新一代网络架构、开放可编程网络
    陈超(1986-),男,浙江湖州人,博士,浙江工商大学副教授、硕士生导师,主要研究方向为下一代无线通信网络技术、网络编码、机器/深度学习等
  • 基金资助:
    国家自然科学基金资助项目(61871468);国家自然科学基金资助项目(61801427);国家自然科学基金资助项目(62111540270);浙江省新型网络标准与应用技术重点实验室基金资助项目(2013E10012);浙江省重点研发计划基金资助项目(2020C01079)

QL-STCT: an intelligent routing convergence method for SDN link failure

Chuanhuang LI, Yangting CHEN, Jingjing TANG, Jiali LOU, Renhua XIE, Chuntao FANG, Weiming WANG, Chao CHEN   

  1. School of Information and Electronic Engineering (Sussex Artificial Intelligence Institute), Zhejiang Gongshang University, Hangzhou 310018, China
  • Revised:2022-01-17 Online:2022-02-25 Published:2022-02-01
  • Supported by:
    The National Natural Science Foundation of China(61871468);The National Natural Science Foundation of China(61801427);The National Natural Science Foundation of China(62111540270);The Key Laboratory of Network Standards and Applied Technology Foundation of Zhejiang Province(2013E10012);The Key Research and Development Program of Zhejiang Province(2020C01079)

摘要:

针对软件定义网络(SDN)链路故障发生时的路由收敛问题,提出了 Q-Learning 子拓扑收敛技术(QL-STCT)实现软件定义网络链路故障时的路由智能收敛。首先,选取网络中的部分节点作为枢纽节点,依据枢纽节点进行枢纽域的划分。然后,以枢纽域为单位构建区域特征,利用特征提出强化学习智能体探索策略来加快强化学习收敛。最后,通过强化学习构建子拓扑网络用于规划备用路径,并保证在周期窗口内备用路径的性能。实验仿真结果表明,所提方法能够有效提高链路故障网络的收敛速度与性能。

关键词: 软件定义网络, 链路故障, 强化学习, 路由收敛

Abstract:

Aiming at the problem of routing convergence when SDN link failure occurs, a Q-Learning sub-topological convergence technique (QL-STCT) was proposed to realize intelligent route convergence when SDN links fail.Firstly, some nodes were selected in the network as hub nodes and divides the hub domains according to the hub nodes, and the regional features were constructed with the hub domain as the unit.Secondly, the reinforcement learning agent exploration strategy was proposed by using the features to accelerate the convergence of reinforcement learning.Finally, a sub-topology network was constructed through reinforcement learning to plan the alternate path and ensure the performance of the alternate path in the periodic window.Experimental simulation results show that the proposed method effectively improves the convergence speed and performance of the link failure network.

Key words: software defined network, link failure, reinforcement learning, routing convergence

中图分类号: 

No Suggested Reading articles found!