Research And Implementation Of Elastic Auto-scaling And Fault Detection Mechanism Of Edge Service

With the rapid development of Internet of Things technology,edge data is increasing explosively.The traditional cloud computing model is unable to meet the key requirements of agile connection,real-time response and flexible expansion of various applications,so the emerging edge computing model arises.Edge computing service platform provides users with core capabilities such as computing,storage and transmission by managing computing units near data sources.And with the development of microservice architecture and container technology,edge computing service system is gradually transitive to a more lightweight,loosely coupled and easily expanded microservice container application architecture,realizing flexible and dynamic adjustment of cluster scale.However,due to the limited heterogeneous resources of edge nodes,the dynamic workload fluctuation and complex network environment,delaysensitive edge services have greater risk of service failure.Existing service fault avoidance methods do not take into account the dependence of microservices and the off-line autonomy of edge nodes,resulting in unreasonable service replica adjustment and inaccurate fault detection,and the reliability of edge services cannot be guaranteed.This paper mainly studies how to guarantee service quality and improve the reliability of edge service system in dynamic edge computing scenarios.Firstly,to solve the problem of dynamic workload fluctuation at edge,this paper proposes a joint optimization strategy of adaptive elastic scaling and service placement considering the resource characteristics of edge nodes and the dependence between microservices.This strategy defines the joint optimization goal as a task delay minimization problem with edge resource and bandwidth constraints,and designs a multi-stage elastic scaling model based on workload prediction of edge microservices and performance evaluation of edge nodes,to dynamically create service replicas and iteratively search for the best placement strategy according to proposed adaptive service placement approach.Secondly,considering the node offline autonomy and other failure scenarios caused by weak edge network environment,this paper proposes a service fault detection method based on multi-party detection model to diagnose the disconnection status of edge nodes with finegrained cloud-edge heartbeat and edge-edge heartbeat mechanism,and a service migration strategy based on weight self-learning is designed to realize fast recovery of edge service faults.Finally,this paper designs and implements the edge computing service management system based on KubeEdge technology,integrating the elastic expansion module and fault detection module,and verifies the system functions and methods.Experiments show that the proposed method can effectively shorten the service response time,avoid service interruption,and improve the reliability of edge service system.