数据与计算发展前沿 ›› 2024, Vol. 6 ›› Issue (5): 91-101.

CSTR: 32002.14.jfdc.CN10-1649/TP.2024.05.009

doi: 10.11871/jfdc.issn.2096-742X.2024.05.009

• • 上一篇    下一篇

多用户多边缘服务器的低碳算网技术研究

沈林江1,*(),崔超1,徐胜霞2,仇树卿1,许俊东1,耿晓巧1   

  1. 1.浪潮通信信息系统有限公司，算力网络研究院，山东 济南 250100
    2.山东信息职业技术学院，电子与通信系，山东 潍坊 261061
  • 收稿日期:2023-01-12 出版日期:2024-10-20 发布日期:2024-10-21
  • 通讯作者: * 沈林江（E-mail: shenlj@inspur.com）
  • 作者简介:沈林江，浪潮通信信息系统有限公司，算力网络研究院院长，公司副总经理，主要从事算力网络相关前沿理论分析、技术研究和产品设计。
    本文中负责系统建设、优化方案设计以及论文总体方向与框架。
    SHEN Linjiang is the director of Computational Power Network Research Institute and deputy general manager of Inspur Communication Information System Co., Ltd. Currently, he is mainly engaged in cutting-edge theoretical analysis, technical research, and product design related to computational power networks.
    In this paper, he is responsible for system and optimization scheme design, and the overall direction and framework of the paper.
    E-mail: shenlj@inspur.com

Research on Low-Carbon Computational Power Network Technology for Multi-User and Multi-Edge Server System

SHEN Linjiang1,*(),CUI Chao1,XU Shengxia2,QIU Shuqing1,XU Jundong1,GENG Xiaoqiao1   

  1. 1. Computational Power Network Research Institute, Inspur Communication Information System Co., Ltd., Jinan, Shandong 250100, China
    2. Department of Electronic Communications, Shandong Vocational College of Information Technology, Weifang, Shandong 261061, China
  • Received:2023-01-12 Online:2024-10-20 Published:2024-10-21

RichHTML

8

PDF

229

摘要：

【目的】算力网络背景下，本文对多用户多服务器的边缘计算系统进行了能耗方面的研究，通过在多个边缘服务器间进行算力任务卸载和动态资源调度，以进一步实现边缘计算系统的能耗优化。【方法】首先建立了包含多用户和多服务器的边缘计算体系与调度系统，分析了影响系统能耗的关键因素，并按照实际算力需求对终端传输功率和边缘服务器频率进行优化，在此基础上设计了多边缘服务器间算力任务动态调度策略，避免由于负载不均导致的能耗偏高问题。【结果】通过理论推导证明了方法的正确性，基于仿真模型和策略设计，验证了本文中方法能够在保障服务质量的同时，实现算力资源间的共享和能耗水平的降低。【局限】本文中方法主要从计算和数据传输两个角度进行分析，对相关模型进行了合理的简化分析，在当前基础上综合考虑网络、应用以及数据安全等维度，并进行落地实践，能够进一步提升方法的应用价值。【结论】基于算力网络对资源的统一纳管和调度能力，在边缘计算等场景中进行合理的资源管理、算力卸载和任务调度，能够有效提升面向业务的服务质量保障和面向底层资源的能耗优化。

关键词: 算力网络, 边缘计算, 能耗优化

Abstract:

[Objective] Under the background of the computational power network, this paper studies the energy consumption of multi-user and multi-server edge computing systems, and further realizes the energy consumption optimization of edge computing systems by offloading computing tasks and dynamic resource scheduling among multiple edge servers. [Methods] First, an edge computing system and scheduling system including multi-user and multi-server are established, the key factors affecting the energy consumption of the system are analyzed, and the terminal transmission power and the clock frequency of the edge server are optimized according to the actual computing power requirements. On this basis, a dynamic scheduling strategy for computing tasks among multiple edge servers is designed to avoid the problem of energy waste caused by the uneven load. [Results] The correctness of the method is proved by theoretical derivation. Based on the simulation model and strategy design, it is verified that the method in this paper can realize the sharing of computing resources and reduce the energy consumption level while ensuring service quality. [Limitations] In this paper, the method is mainly analyzed from the perspectives of calculation and data transmission, the relevant model is reasonably simplified and analyzed, and the dimensions of network, application and data security are comprehensively considered on the current basis, and the implementation practice is carried out, which can further improve the application value of the method. [Conclusions] Based on the unified management and scheduling capabilities of resources in the computational power network, reasonable resource management, computing task offloading, and task scheduling are carried out in edge computing and other scenarios, which can effectively improve service-oriented QoS assurance and energy consumption optimization for underlying resources.

Key words: computational power network, edge computing, energy-efficiency optimization