摘要
随着互联网的发展,需要对IPv6网络进行性能提升,软件定义网络可以通过其特有的段路由架构对IPv6进行改进。将现有的IPv6网络直接升级为完整的SRv6网络是非常困难的。因此,提出一种基于部分部署的SRv6的网络优化算法。通过对分散的SRv6节点中的TE算法进行权重调整,将TE问题转化为深度强化学习问题,优化OSPF权重、SRv6节点部署和流量路径。实验结果表明,提出的优化算法,在不同数量的节点下,可以获得与完整SR网络相同的性能,有着满足实际需求,具有较好性能。
With the development of the Internet,it is necessary to improve the performance of IPv6 networks.Software-defined networks can improve IPv6 through its unique segment routing architecture.However,it is very difficult to directly upgrade an existing IPv6 network to a complete SRv6 network.Therefore,a network optimization algorithm based on partially deployed SRv6 was proposed.The weight of the TE algorithm in the scattered SRv6 nodes was adjusted to transform the TE problem into a deep reinforcement learning problem and optimize the OSPF weights,SRv6 node deployment and traffic path.Experimental results show that the proposed optimization algorithm can achieve the same performance as a complete SR network under a diffe-rent number of nodes,and it can meet actual needs with better performance.
作者
刘威
黄萍
孙凤杰
LIU Wei;HUANG Ping;SUN Feng-jie(Information Center,Shenzhen Power Supply Bureau Limited Company,Shenzhen 518000,China;School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China)
出处
《计算机工程与设计》
北大核心
2022年第4期930-940,共11页
Computer Engineering and Design
基金
国家自然科学基金面上基金项目(51677065)
中央高校基本科研业务费专项基金项目(2016MS06)。
关键词
段路由
深度强化学习
软件定义网络
IPV6
网络优化
segment routing
deep reinforcement learning
software-defined network
IPv6
network optimization
作者简介
刘威(1988-),男,河南许昌人,硕士,高级工程师,研究方向为电力信息;黄萍(1993-),女,广东韶关人,助理工程师,研究方向为电力信息;孙凤杰(1964-),男,北京人,硕士,教授,研究方向为通信与信息系统、光伏组件监测。E-mail:wuchu295@163.com。
