Design And Implementation Of Dynamic Pod Scaling Scheduling Platform Optimized For Microservice Deployment

With the rapid development of information technology and the continuous emergence of technological innovations,various industries are facing new challenges brought by increasing business demands.This has led to increasingly complex software systems,larger volume of information projects,cumbersome business processes and the high costs of management,operation and maintenance.In this context,microservices effectively solve the challenges of centralization,high cost,and difficult maintenance of the traditional architecture.However,the digital transformation of the government and the rapid development of enterprise businesses on the cloud require both online and continuously running microservices to ensure critical business and the optimal utilization of resources while maintaining high availability,which requires a microservice to be able to implement auto scaling and flexible deployment of microservices to achieve more efficient management and on-demand expansion.Focusing on the above problems,this paper designs and implements a dynamic Pod scaling scheduling platform based on the load and performance requirements of microservices,which controls the number of Pods through load monitor and load predict,so as to realize the auto scaling and flexible deployment of microservices,and provide a high-availability and high QoS operating environment.Based on the design idea of "platform plus service",this paper provides four functional modules of microservices,including containerized deployment,load monitor and predict,Pod auto scaling,and node information.The specific work is as follows.First,the platform uses the Helm package management tool for containerized deployment of microservices,encapsulates user microservices,packages and ports applications and services to containers,so as to achieve flexible deployment and management of virtualization and applications.It mainly includes production,installation and instance management of container templates,which simplifies application deployment version control,deletion and update,package and release,achieves rapid deployment,and reduces the complexity of development and operation.Second,the platform uses the Attention-based GRU algorithm to predict the workload of the next cycle by determining the load predict indicators,so as to formulate the horizontal elastic scaling strategy in a targeted manner.At the same time,the platform monitors data such as microservice instances and users’ microservice load indicators from different time dimensions,providing users with a clear monitoring platform,thereby enabling high available,efficient and high QoS platform services.Third,with the horizontal auto scaling strategy based on load predict,the platform plans the number of copies of the target Pod in advance,increases or decreases the number of Pod copies,and achieves auto scaling through dynamic scheduling of Pod resources.Similarly,according to the differences in user requirements,the number of Pods can be scaled automatically or manually,so as to provide flexible support for microservices when business requirements change.Fourth,the platform obtains the node list page and details page through the node information module to view information such as node status,Pod list and resource usage,which facilitates online operations to access the computing resources of the cluster.Based on the monitor and predict of load indicators provided by Pods for microservices,this paper further implements the flexible deployment of microservices and the auto scaling of Pods,thereby enhancing the efficiency of system development and management,reducing system maintenance costs and improving system flexibility and agility.It optimizes the service levels and achieves high availability and high QoS to provide users with a reliable service environment for continuous operation.