Research And System Implementation Of Resource Scheduling Method For Edge Nodes In Edge Computing

With the maturity of 5G technology and the development of emerging industries such as AR / VR and autonomous driving,centralized cloud computing is increasingly difficult to meet the public's demand for latency,making the computing paradigm shift towards edge computing.However,due to the characteristics of distributed multi-deployment,edge computing cannot guarantee that each edge node cluster has sufficient resources.Faced with the limited resources of edge node clusters,how to provide users with high-quality,low-latency services through reasonable resource scheduling has become a problem worthy of research.To solve this problem,this paper studies the resource scheduling method based on edge computing.Based on the theory of linear programming and stable matching,server quantity optimization model based on linear programming and the improved stable matching algorithm adapted to edge computing environment are proposed to improve the transmission delay and service quality of user tasks.At the same time,this paper also realizes the resource management business function in the edge computing system as the carrier of the above two algorithms.Specifically,the main work of this paper is as follows:(1)Propose a server quantity optimization model based on linear programmingAiming at the problem of resource scheduling in the scenario where edge computing resources are limited,this paper studies how to minimize the number of servers and deploy more clusters of edge nodes to reduce user task transmission delays.In order to solve this problem,this paper introduces the NP-hard multi-knapsack problem.The original multi-knapsack problem that maximizes the value of items is transformed into a resource scheduling problem that minimizes the number of servers based on the mathematical model of resource construction in this paper.A server quantity optimization model based on linear programming is proposed.This article starts with the two indicators of server usage and server resource occupation.This paper starts with two indexes of server usage and server resource occupation,and compares the resource scheduling results of this model obtained by implicit enumeration method and the resource scheduling results obtained by the first adaptation algorithm on the kubernetes platform.The results show that the server quantity optimization model based on linear programming can effectively reduce the number of servers and resource consumption.(2)Propose a improved stable matching algorithm adapted to edge computing environmentAiming at the problem of resource scheduling in scenarios where edge computing resources are not constrained,study how to rationally configure the mapping relationship between virtual resources and physical resources to schedule virtual resource units to the most appropriate server to improve user service quality.In order to solve this problem,this paper introduces a container stable matching algorithm,which is applied to the above scenarios and optimized according to the actual situation.An improved stable matching algorithm adapted to edge computing environment is proposed.This article starts with the resource scheduling time index,and compares this algorithm with the polling scheduling algorithm on the Kubernetes platform.The results show that the stable matching algorithm adapted to edge computing environment can reduce the resource scheduling time by about 8%.(3)Design and implement the resource management business function in the edge computing systemIn order to realize the server quantity optimization model based on linear programming and the improved stable matching algorithm adapted to edge computing environment,this paper designs and implements the resource management business function in the edge computing system.This business function is based on the Kubernetes platform,which manages pods,mirrors,and data volumes,and implements the resource scheduling function including the above algorithms.At the same time,considering the real needs of resource scheduling,this article also implements the user management module and resource monitoring module to divide and monitor resources,and performs functional verification on these three modules.