面向科学计算的算力网络技术发展展望  

Development and Prospects of Computing Power Network for Scientific Computing

作　　者：周旭[1] 李琢 范鹏飞 李菁菁[1] Zhou Xu;Li Zhuo;Fan Pengfei;Li Jingjing(Computer Network Information Center,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]中国科学院计算机网络信息中心,北京100190

出　　处：《信息通信技术》2023年第1期37-43,共7页Information and communications Technologies

摘　　要：随着科学计算时代的到来,海量科学数据的产生带来丰富的基础性科技资源,也为计算和传输带来巨大的压力,需要大量的资源来处理,亟需算网融合技术将算力和网络有机结合。算力网络连通分散的、碎片化的算力资源,构建一体化的ICT基础设施,提供计算能力和网络能力等服务。文章研究面向科学计算的算力网络技术,搭建算力网络总体架构,采用资源度量、资源调度和算力路由等技术,部署算力网络原型系统,提供高性能的科学计算能力,完成复杂模拟和科学计算。在高能物理、天文领域等典型应用开展试验示范,能够显著提升科学大数据的网络传输效率和大数据处理能力,提升科研信息化服务能力与水平,加速重大创新成果产出。With the advent of the era of scientific computing,the generation of massive scientific data has brought rich fundamental technological resources,as well as enormous pressure on computation and transmission,which requires a large amount of resources to handle.There is an urgent need for the integration of computing and networking through computingnetwork fusion technology.The computing power network connects dispersed and fragmented computing resources to build an integrated ICT infrastructure,providing services such as computing power and network capabilities.This article researches on computing power network technology for scientific computing,construction of the overall architecture of the computing power network,adoption of technologies such as resource measurement,resource scheduling,and computing power routing,deployment of prototype systems for computing power networks,and provision of high-performance scientific computing capabilities to complete complex simulations and scientific computing.Conducting experimental demonstrations in typical applications such as high-energy physics and astronomy can significantly improve the network transmission efficiency and big data processing capabilities of scientific big data,enhance the level and capacity of scientific research information services,and accelerate the output of major innovative achievements.

关 键 词：算力网络 科学计算 计算需求 

分 类 号：TP30[自动化与计算机技术—计算机系统结构]