信息通信技术与政策

信息通信技术与政策

信息通信技术与政策 ›› 2023, Vol. 49 ›› Issue (5): 42-47.doi: 10.12267/j.issn.2096-5931.2023.05.007

专题:算网融合 上一篇    下一篇

面向异构算力网络调度的研究

Research on heterogeneous computing power network scheduling

张慷1,2, 桑洁丽1, 吴美希3   

  1. 1.中国电信股份有限公司上海分公司,上海 200120
    2.天翼云科技有限公司上海分公司,上海 200085
    3.中国信息通信研究院云计算与大数据研究所,北京 100191
  • 收稿日期:2023-04-13 出版日期:2023-05-25 发布日期:2023-05-30
  • 作者简介:
    张慷,中国电信股份有限公司上海分公司资深经理、总师室主任,天翼云科技有限公司上海分公司副总经理,教授级高工,主要从事IP、云计算、系统集成等方面的研究工作
    桑洁丽,中国电信股份有限公司上海分公司天翼云能力运营中心产品运营中心经理,工程师,主要从事算力调度、云网融合、多云管理等方面的研究工作
    吴美希,中国信息通信研究院云计算与大数据研究所数据中心部副主任,工程师,主要从事数据中心相关的政策支撑、产业咨询、技术研究和标准制定等方面的研究工作

ZHANG Kang1,2, SANG Jieli1, WU Meixi3   

  1. 1. China Telecom Corporation Limited Shanghai Branch, Shanghai 200120, China
    2. China Telecom Cloud Science & Technology Limited Shanghai Branch, Shanghai 200085, China
    3. Cloud Computing & Big Data Research Institute, China Academy of Information and Communications Technology, Beijing 100191, China
  • Received:2023-04-13 Online:2023-05-25 Published:2023-05-30

摘要:

在“东数西算”工程背景下,算力基础设施迎来了飞速发展,产生了大量异构算力资源。云—边—端的算力与网络协同,能够满足多种异构算力资源与众多应用场景的高效匹配。基于异构算力网络调度展开研究,为异构算力网络融合调度技术提供研究思路,系统性地分析了技术发展现状、算力网络融合调度关键技术与挑战,提出了众多应用场景。但实际应用中异构算力网络调度更为复杂,仍面临许多不确定性挑战,需要不断完善。

关键词: 异构算力, 算力网络, 高效调度, 算力应用

Abstract:

Under the background of the project of national computing network to synergize east and west, the computing power infrastructure has ushered in rapid development, resulting in a large number of heterogeneous computing power resources. The computing power and network collaboration of cloud, edge, and device can meet the efficient matching of multiple heterogeneous computing power resources with many application scenarios. Based on the research of heterogeneous computing power network scheduling, this paper provides a research idea for the scheduling technology to achieve heterogeneous computing and network convergence, systematically analyzes the technological development status, key technologies and challenges, and points out many application scenarios. However, in practical applications, heterogeneous computing network scheduling is more complex, and it still faces many uncertainties and challenges, which requires continuous improvement of technology.

Key words: heterogeneous computing power, computing power network, efficient scheduling, computing power application

中图分类号: