面向算力制造全流程的企业园区算力网络系统和关键技术  

Computing Power Network System and Key Technologies in Enterprise Campus for the Entire Procedure of Computing Power Manufacturing

作　　者：章璐 史庭祥[1,2] 张健 Zhang Lu;Shi Tingxiang;Zhang Jian(State Key Laboratory of Mobile Network and Mobile Multimedia Technology,Shenzhen 518055,China;ZTE Corporation,Nanjing 210012,China)

机构地区：[1]移动网络和移动多媒体技术国家重点实验室,深圳518055 [2]中兴通讯股份有限公司,南京210012

出　　处：《信息通信技术》2024年第4期57-63,共7页Information and communications Technologies

基　　金：国家重点研发计划基金资助项目(No.2021YFB2900200)。

摘　　要：数字经济催生算力需求突飞猛进,企业数字化转型依赖算力资源和相关应用。围绕算力制造全流程解析的四大类算力应用,提出企业园区算力网络的系统架构,并基于此构建覆盖研发云、办公云、生产云和终端云的四个子系统,进而建立围绕三大技术创新的企业园区算力网络系统。文章还首次提出算力制造的核心竞争力问题等价于这些应用使用算力的效率和成本问题,而唯有在资源和运营上实施双重定制,才能实现资源利用的极致效率和运营服务的极致体验,从而达成算力制造的降本提效目标。该系统和关键技术在运营商和企业合作项目开展试点验证,结果表明,在改善应用体验和算力资源效率上取得显著成效,成为算力服务器竞争力的关键因素。The demand for computing power is growing rapidly in the digital economy,and digital transformation of enterprises depends on computing power resources and related applications.Based on the analysis of the four major categories of computing power applications in the entire process of computing power manufacturing,a system architecture for computing power networks of enterprise campus is proposed.Based on this,four subsystems covering research and development cloud,office cloud,production cloud,and terminal cloud are constructed,and a computing power network system for enterprise campus is established around the three major technological innovations.This paper also proposes for the first time that the core competitiveness of computing power manufacturing is equivalent to the efficiency and cost of using computing power in these applications.Only by implementing dual customization in resources and operations can we achieve the ultimate efficiency of resource utilization and the ultimate experience of operational services,thereby achieving the goal of cost reduction and efficiency improvement in computing power manufacturing.The system and key technologies were pilot verified in the cooperation projects between operators and enterprises.The results show that significant achievements have been made in improving the application experience and computing resource efficiency,becoming a key factor in the competitiveness of computing servers.

关 键 词：算力制造 算力调度 算力并网 云边端协同 算力路由 企业广域网 业务优先级调度 业务服务质量 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]