摘要
【目的】进入算力时代以来,伴随泛在接入、万物互联,全社会数据量迎来爆发式增长。需要通过算力网络解决大数据计算资源不足、异构算力不足、边缘算力不足等问题。【方法】基于算力网络,重新设计大数据计算体系架构,通过资源编排层提供大数据算力编排、资源封装、统一调度能力。同时利用粒子群算法,智能计算和寻找出资源调度分配的最优节点,使资源分配达到最优平衡。【结果】通过新型的资源调度方法,智能调度各类算力,从根本上解决大数据计算资源不足、异构算力不足、边缘算力不足等问题。【结论】使用基于算力网络的大数据计算资源智能调度分配方法,能够智能调度社会闲散算力、异构算力和边缘算力,从全局角度解决算力需求不足和算力分布不均等问题。
[Objective]Since entering the age of computing power,with ubiquitous access and the interconnection of everything,the amount of data in the whole society has ushered in explosive growth.We need to solve the problems of insufficient big data computing resources,heterogeneous computing power,and edge computing power through the computing power network.[Methods]Based on the computing power network,redesign the big data computing architecture,and provide the ability of big data computing power scheduling,resource encapsulation,and unified scheduling through the resource scheduling layer.At the same time,the particle swarm optimization algorithm is used to intelligently calculate and find the optimal node in resource scheduling and allocation,so as to achieve the optimal balance of resource allocation.[Results]Through the new resource scheduling method,intelligent scheduling of all kinds of computing power can fundamentally solve the problems of insufficient big data computing resources,heterogeneous computing power,and edge computing power.[Conclusions]The intelligent scheduling and allocation method of big data computing resources based on computing power networks can intelligently schedule social idle computing power,heterogeneous computing power,and edge computing power,and solve the problem of insufficient computing power demand and uneven distribution of computing power from a global perspective.
作者
金天骄
栗蔚
JIN Tianjiao;LI Wei(China Mobile Group Zhejiang Co,Ltd,Hangzhou,Zhejiang 310030,China;China Academy of Information and Communications Technology,Beijing 100191,China)
出处
《数据与计算发展前沿》
CSCD
2022年第6期29-37,共9页
Frontiers of Data & Computing
关键词
算力网络
大数据计算
资源调度
computing power network
big data computing
resource scheduling
