摘要
5G与边缘计算等应用场景的兴起,使得计算、存储等基础IT资源离散部署与多级部署成为一大发展趋势,如何结合网络基础设施为用户提供更为便捷的、定制化的专属服务,提高整体资源利用率,成为网络技术发展的新方向。重点分析了网络虚拟化的特征与发展、资源发展与融合趋势以及新兴业务对资源供给新需求等,提出了一种基于多维资源融合化的网络虚拟化架构,即通过在网络虚拟化架构中引入资源发现与资源交易过程,将多方、异构的资源有机整合成统一资源平面,在此基础上根据用户需求进行资源切割与虚拟化,以适应新兴业务对资源需求快速灵活的特点,并能够有效提高资源利用效率,降低运营成本。由于资源联合优化涉及多个领域的研究,网络虚拟化架构将随着业务需求与商业模式的发展而不断完善,以期为未来业务发挥重要作用。
With the development of 5G and edge computing,the multi-level and distributed deployment of basic IT resources such as computing resources and storage resources becomes mainstream.The way to provide users with more convenient and customized services to improve resource utilization becomes a new interest in network.The characteristics and development of the network virtualization,interworking trend and development of various resources,as well as the new requirements of emerging services for resource supply were analyzed.On the basis of the analysis,a network virtualization architecture based on multi-dimensional resource interworking was proposed.That is,multi-party and heterogeneous resources were integrated into a unified resource plane by introducing the process of resource discovery and resource transaction,and then resource slicing and virtualization were performed according to user requirements to meet the flexible resource requirements of emerging services.Resource utilization efficiency was improved effectively and operating cost was reduced.Since the joint optimization of resources involves multiple research fields,the network virtualization architecture will be constantly improved with the development of user requirements and business models,playing an important role for future services.
作者
雷波
王江龙
赵倩颖
余勇志
杨明川
LEI Bo;WANG Jianglong;ZHAO Qianying;YU Yongzhi;YANG Mingchuan(Research Institute of China Telecom Co.,Ltd.,Beijing 102209,China;Chongqing Branch of China Telecom Co.,Ltd.,Chongqing 404100,China)
出处
《电信科学》
2020年第7期42-54,共13页
Telecommunications Science
基金
国家重点研发计划基金资助项目(No.2018YFB1800100)。
关键词
网络虚拟化
边缘计算
算力网络
network virtualization
edge computing
computing power network
