Implementation Of Microservice Architecture And Optimization Based Kubernetes

Kubernetes,which treats containers as services and provides automatic restart,elastic scaling and other functions,is the mainstream platform for deploying microservices today.However,kubernetes cluster scheduling is a mechanism carried out on static resources in default implementation.In the architectural deployment of practical application scenarios of e-commerce,there are still disadvantages such as redundant calculation on expansion capacity and insufficient utilization of node resources,which easily lead to bottlenecks in cluster concurrency performance and high availability.Aiming at the above challenges,based on Kubernetes as the deployment framework of microservice architecture,this paper optimized the default scheduling of traditional e-commerce microservice architecture.The paper divides the whole architecture into pre-service,business service and basic service.Using technologies such as K8S-nacos clustering,o Auth2(Open Authorization),and OSS(Object Storage)to address the common challenges faced by microservice architectures such as security performance and reliability.Through DDD(Domain Driven Design),the service division of e-commerce system is carried out,and the business service part and the pre-service part are realized,so as to provide a practical basis for scheduling optimization.At the same time,the paper analyzes the source code of Kubernetes’ default scheduling,improves the default scheduling process and scheduling algorithm,and realizes the custom scheduler.Custom scheduler by setting the CPU,memory,disk capacity index and network bandwidth resources,and the resource index of time series to predict the demand for replication of resources value,at the same time,according to each Node in the cluster resources available to calculate the different resources tend to be the exclusive use of a pod of heavy,use scheduling strategy based on weight self-learning Node with the pod binding.After the custom scheduler is implemented,the Deployment file is created and its image is deployed to the Kubernetes cluster,and when the Pod is created in the cluster,the scheduler used by the pod is specified as the custom scheduler.The above scheme is used to design the micro-service e-commerce system in combination with DDD thinking,so as to provide practical basis for the feasibility of the architecture deployment on Kubernetes platform.At the same time,it is compared with the default scheduling policy in terms of processing concurrency,capacity expansion replica creation time and actual resource utilization.The experimental results show that using the customized scheduling strategy can better satisfy the reliability of the microservice architecture and improve the load performance and resource utilization of the entire architecture cluster.