边缘算力网络中智能算力感知路由分配策略研究  

Study on Intelligent Computing Aware Route Allocation Policy in Edge Computing-Aware Networks

作　　者：孙钰坤 张兴[1] 雷波[2] SUN Yukun;ZHANG Xing;LEI Bo(Key Laboratory of Universal Wireless Communications,Ministry of Education,Beijing University of Posts and Telecommunications,Beijing 100876,China;Research Institute of China Telecom Co.,Ltd.,Beijing 102209,China)

机构地区：[1]北京邮电大学泛网无线通信教育部重点实验室,北京100876 [2]中国电信股份有限公司研究院,北京102209

出　　处：《无线电通信技术》2022年第1期60-67,共8页Radio Communications Technology

基　　金：之江实验室开放课题(2020LCOAB01);国家自然科学基金(62071063)。

摘　　要：增强现实、自动驾驶、智慧城市、工业互联网等新型业务应用对网络算力的需求逐渐增强,然而,边缘算力网络系统面临着网络共存的问题——负载不均衡,导致一部分边缘服务器无法满足业务应用的处理需求,另一部分边缘服务器的算力资源处于空闲状态。为了高效协同地感知利用泛在、异构的算力资源,提升6G通信网络的内生感知和算力自适应能力,急需对边缘算力网络中任务路由策略以及算力资源分配进行研究。首先介绍了面向6G网络愿景的算力感知网络的演进和需求,然后构建了任务调度智能决策、存储资源和算力资源按需分配的联合优化路由控制与资源分配的智能任务调度模型,最后提出了一种基于Floyd算法的算力感知路由调度策略解决智能任务调度问题。仿真结果表明,与就近调度固定分配资源策略相比,所提的基于Floyd算法的算力感知路由调度策略可以让更多的用户受益,缩短用户业务的平均处理时延,提高边缘算力网络中存储资源和计算资源的利用率。Emerging network services and applications such as augmented reality,self-driving,smart cities,industrial Internet,put forward higher requirement for computing power.However,computing-aware network is faced with the problem which traditional network is also faced with,i.g,load unbalance.Some edge servers fail to provide enough resources,but the resources of the others are under-utilized.To aware and leverage ubiquitous and heterogeneous computing resource efficiently and collaboratively,and improve 6G communication network endogenous perception and computing adaptive ability,it is urgent to break through computing-aware networking route policy and resource allocation.First,the evolution and demand of computing-aware networking for the vision of 6G networks are analyzed.Then,an intelligent task scheduling model for joint routing control and resource allocation is constructed,which makes intelligent decisions for task scheduling and resources on demand.Finally,the computing aware-route-allocate policy based on Floyd algorithm is put forward.Numerical simulation results show that the computing aware-route-allocate policy based on Floyd algorithm can satisfy more users,shorten average delay for task process,and increase the utilization ratio of storage and computing resources.

关 键 词：算力感知网络 任务调度 融合资源分配 FLOYD算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]