数据与计算发展前沿 ›› 2022, Vol. 4 ›› Issue (5): 11-22.

CSTR: 32002.14.jfdc.CN10-1649/TP.2022.05.002

doi: 10.11871/jfdc.issn.2096-742X.2022.05.002

• 专刊:东数西算:开启算力经济时代的世纪工程(上) • 上一篇    下一篇

基于互联云及多云的云化基础设施算力调度

程莹*(),裴小燕   

  1. 中国联合网络通信集团有限公司,北京 100033
  • 收稿日期:2022-08-01 出版日期:2022-10-20 发布日期:2022-10-27
  • 通讯作者: 程莹
  • 作者简介:程莹,中国联合网络通信集团有限公司科技创新部,国际标准与开源项目主管,高级工程师,同时担任ITU-T SG11 Q4报告人、ITU-T SG13 Q19报告人以及CCSA TC1 WG5副组长。主要研究方向为云计算、云网融合、网络软化、人工智能等。
    本文主要承担工作为全文的整体设计及撰写,云网融合相关国际标准进展分析、基于多云及互联云的云化基础设施算力调度参考架构的提出及示例设计。
    CHENG Ying is a senior engineer and supervisor on interna-tional standards and open source projects in Science & Techno-logy Innovation Department, China Unicom, and serves as Ra-pporteur of ITU-T SG11 Q4, Rapporteur of ITU-T SG13 Q19, and deputy team leader of CCSA TC1 WG5. Her research interests include cloud computing, cloud and network convergence, network softwarization, artificial intelligence, etc.
    In this paper, she is responsible for overall design and writing, international standards progress analysis on cloud and network convergence, and reference architecture and application ex-ample design on inter-cloud and multi-cloud-based computing power scheduling of cloud-enabled infrastructure. E-mail: chengying10@chinaunicom.cn

Computing Power Scheduling of Cloud-Enabled Infrastructure Based on Inter-Cloud and Multi-Cloud

CHENG Ying*(),PEI Xiaoyan   

  1. China Unicom, Beijing 100033, China
  • Received:2022-08-01 Online:2022-10-20 Published:2022-10-27
  • Contact: CHENG Ying

RichHTML

0

PDF

356

摘要：

【目的】“东数西算”工程从国家层面启动了全国一体化大数据中心体系的规划和建设,跨算力服务商的大规模算力服务调度成为重要的研究议题。【应用背景】本文适用于多个大型云服务商协同提供组合算力服务。【方法】本文提出了基于多云及互联云的云化基础设施算力调度参考架构并给出了相关关键技术及应用示例。【结果】本参考架构可以基于服务商合作关系的数据模型,实现屏蔽异构云化基础设施算力资源的自动化部署。【局限】真实场景中的云际互联关系更为复杂,数据模型的设计仍面临许多挑战。【结论】基于多云及互联云的云化基础设施算力调度可以作为“东数西算”全国一体化算力调度的重要技术选择之一。

关键词: 云网融合, 算力调度, 多云, 互联云, 不可变基础设施, 基础设施即代码

Abstract:

[Objective] The “East-West Computing Requirement Transfer” project has initiated the planning and construction of the integrated big data center system at the national level. Scheduling of large-scale computing services across different computing service providers has become an important research topic. [Application background] This paper applies to multiple large cloud service providers collaborating to provide combined computing power services. [Methods] This paper proposes a reference architecture for computing power scheduling of cloud-enabled infrastructure based on multi-cloud and inter-cloud, and presents related key technologies and application examples. [Results] This reference architecture can automate the deployment of shielded heterogeneous computing power resources of cloud-enabled infrastructure based on the data model of the service providers’ partnership. [Limitations] The service providers’ interconnection relationships in real-world scenarios are more complex, and the design of data models still faces many challenges. [Conclusions] The inter-cloud and multi-cloud-based computing power scheduling of cloud-enabled infrastructure can be one of the important technology choices for the national integrated computing power scheduling under the “East-West Computing Requirement Transfer” strategy.

Key words: cloud and network convergence, computing power scheduling, multi-cloud, inter-cloud, immutable infrastructure, infrastructure as code