支持多模态网络的可扩展异构服务功能链并行编排部署系统  

Parallel orchestration and deployment system for scalable heterogeneous service function chain supporting polymorphic network

作　　者：陈浩 杨芫[1,2] 徐明伟 裴丹[1,2] 尤艺霖 CHEN Hao;YANG Yuan;XU Mingwei;PEI Dan;YOU Yilin(Department of Computer Science and Technology,Tsinghua University,Beijing 100084,China;Beijing National Research Center for Information Science and Technology(BNRist),Beijing 100084,China;Zhongguancun Laboratory,Beijing 100083,China;Institute for Network Sciences and Cyberspace,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学计算机科学与技术系,北京100084 [2]北京信息科学与技术国家研究中心,北京100084 [3]中关村实验室,北京100083 [4]清华大学网络科学与网络空间研究院,北京100084

出　　处：《通信学报》2022年第9期1-11,共11页Journal on Communications

基　　金：国家重点研发计划基金资助项目(No.2019YFB1802504);国家自然科学基金资助项目(No.61872209,No.61832013)。

摘　　要：针对如何在万台服务器规模的多模态数据中心网络内,实现秒级的服务功能链扩容编排部署问题,提出一种可扩展的异构服务功能链并行编排部署(SHOD)系统。SHOD系统对数据中心进行分区,并采用资源使用量最小化的并行编排器设计,从而将服务功能链扩容编排时间降低至秒级。SHOD系统采用满足单一权职原则的中介者设计模式,实现了对异构设备的可扩展的编排部署和服务功能链建链。实验结果表明,SHOD系统的扩容编排时间下降至Daisy的1/12,同时仅引入1.5倍的CPU占用率开销。Aiming at the problem of second-level service function chain orchestration and deployment in the polymorphic data center with the scale of 10 000 servers, a scalable heterogeneous SFC parallel orchestration and deployment(SHOD) system was proposed. The scale-out orchestration time of service chain was reduced to seconds by partitioning the data center and adopting a parallel orchestrator design that minimizes resource usage. Using a mediator design model that satisfies the single-responsibility principle, the heterogeneous equipment scalable orchestration and deployment were realized as well as service function chain building chain. Experimental results show that in a data center with 10 000 servers,the scale-out orchestration time of SHOD system is reduced to 1/12 of the existing system-Daisy, while only 1.5 times CPU utilization overhead is introduced.

关 键 词：多模态网络 网络功能虚拟化 服务功能链 可扩展性 编排部署系统 

分 类 号：TN393[电子电信—物理电子学]