基于网络切片的移动雾计算系统资源调度策略  

Resource Scheduling Strategy of Mobile Fog Computing System Based on Network Slicing

作　　者：张小佳 文鹏 唐胜达 ZHANG Xiao-jia;WEN Peng;TANG Sheng-da(School of Mathematics and Statistics,Guangxi Normal University,Guilin 541006,China;School of Science and Technology,Guilin University,Guilin 541006,China)

机构地区：[1]广西师范大学数学与统计学院,广西桂林541006 [2]桂林学院理工学院,广西桂林541006

出　　处：《长春师范大学学报》2022年第8期17-25,共9页Journal of Changchun Normal University

摘　　要：针对基于网络切片的移动雾计算系统资源调度策略进行研究.系统根据切片的业务需求,制定各类业务的切片,将网络切片划分优先级,MFC系统中移动用户的计算请求接入对应的网络切片,并为其分配一定的资源以满足基本需求.本文在基于网络切片的MFC系统中提出了一种在系统虚拟机资源池中进行资源分配的最优方案.以最大化MFC系统的长期期望奖励.将优化问题表述为具有定义状态空间、动作空间、奖励模型和转移概率分布的无限视界半马尔可夫决策过程.通过迭代算法求解SMDP问题,得到了相应的最优方案.本文可以提供指导雾计算系统下资源最优分配的方案,有助于决定移动用户请求应该分配多少资源.数值结果表明,该方案优于其他资源分配方案,提高了系统奖励.This paper studies the resource scheduling strategy of mobile fog computing system based on network slicing.First of all,the system formulates various service slices according to the service requirements of slices.Specifically,the network slices are prioritized.The computing requests of mobile users in the MFC system access the corresponding network slices,and certain resources are allocated to them to meet the basic needs.This paper proposes an optimal scheme for resource allocation in the system virtual machine resource pool in the MFC system based on network slicing to maximize the long-term expected reward of the MFC system.Secondly,the optimization problem is formulated as an infinite horizon semi-Markov decision process with a defined state space,action space,reward model,and transition probability distribution.Finally,an iterative algorithm is used to solve the SMDP problem,and the corresponding optimal solution is obtained.The proposal in this paper can provide guidelines for fog computing to help determine how much resources should be allocated for mobile user requests.Numerical results show that this scheme outperforms other resource allocation schemes,improving system rewards and the experience gained by mobile users.

关 键 词：网络切片 移动雾计算系统 半马尔可夫决策过程 VM分配 

分 类 号：O211.6[理学—概率论与数理统计]