数据与计算发展前沿 ›› 2025, Vol. 7 ›› Issue (2): 12-21.

CSTR: 32002.14.jfdc.CN10-1649/TP.2025.02.002

doi: 10.11871/jfdc.issn.2096-742X.2025.02.002

• 专刊：中国科技云10周年 • 上一篇    下一篇

算力网络资源的统一建模及标识研究

吕航*(),邢文娟,马小婷   

  1. 中国电信股份有限公司研究院，北京 102209
  • 收稿日期:2025-02-12 出版日期:2025-04-20 发布日期:2025-04-23
  • 通讯作者: 吕航
  • 作者简介:吕航，中国电信股份有限公司研究院，高级工程师，主要研究领域为云网融合、算力网络、区块链、5G技术等。
    本文中负责论文的综述、资源统一建模设计、标识编码结构设计等内容，并撰写了初稿。
    LYU Hang is a senior engineer at the China Telecom Research Institute. His main research areas are cloud network convergence, computing networks, blockchain, and 5G technology.
    In this paper, he is responsible for the review of the paper, unified modeling and design of resources, identification and coding structure design, and writing the initial draft.
    E-mail: lvhang@chinatelecom.cn
  • 基金资助:
    国家重点研发计划(2022YFB2901405)

Research on Unified Modeling and Identification of Computing Power Network Resources

LYU Hang*(),XING Wenjuan,MA Xiaoting   

  1. Research Institute of China Telecom Co., Ltd., Beijing 102209, China
  • Received:2025-02-12 Online:2025-04-20 Published:2025-04-23
  • Contact: LYU Hang

RichHTML

1

PDF

13

摘要：

【目的】随着新业态、新模式对算力网络要求的不断提升，资源协同编排调度与交易机制面临新的挑战。本文旨在解决算力网络资源标识的关键问题，构建可支撑资源交易与协同调度的系统性方法。【方法】首先设计资源抽象描述框架，通过属性解耦实现多维度资源表征，其次提出基于多维特征的算力资源统一建模方法，建立资源分类体系；并在此基础上构建层次化资源度量模型；进而创新性地提出算力网络资源标识的生成算法；最后面向实际应用场景，研究设计算力资源交易机制和动态优先级编排调度策略。【结论】研究成果有效解决了算力资源异构性导致的算力网络运行低效的问题，提出的标识体系为资源确权交易及协同调度提供了技术支撑，为构建智能化的算力网络生态系统提供了理论和方法基础。

关键词: 算力网络, 资源度量, 算力网络资源标识, 运行机制

Abstract:

[Objective] With the continuously increased requirement to computing power network by new commercial activities and models, resource collaborative scheduling and transaction mechanisms are facing new challenges. This article aims to address the key issues of resource identification in computing power networks and build a systematic approach that can support resource transactions and collaborative scheduling. [Methods] Firstly, a resource abstract description framework is designed to achieve multidimensional resource representation through attribute decoupling. Secondly, a unified modeling method for computing power resources based on multidimensional features is proposed to establish a resource classification system. Building upon this, a hierarchical resource measurement model is developed. Furthermore, an innovative algorithm for generating computing power network resources indicators is proposed. Finally, based on practical application scenarios, a transaction mechanism for computing power resources and a dynamic priority scheduling strategy are designed. [Conclusions] The research effectively solve the problem of inefficient operation of computing power networks caused by the heterogeneity of computing power resources. The proposed identification system provides technical support for resource ownership transactions and collaborative scheduling, which paves a theoretical and methodological foundation for building an intelligent computing power network ecosystem.

Key words: computing power network, resource measurement, computing network resource identification, operating mechanism