Research And Implementation Of Dynamic Scheduling Of Container Cloud Platform Resources Based On Kubernetes

Cloud computing is a service that allows dynamic rental of computing resources on demand.The flexibility of cloud resources allows for the perfect use of computing resources,because the number of resources can be adjusted according to actual needs.Nowadays,with the increasing popularity of cloud platforms,how to provide the resources in cloud platforms,including CPU,memory,GPU and network,to users more accurately is a problem to be solved.With the maturity and popularity of container technology represented by Docker,people have new ideas for resource management of container cloud platform.The industry's vision has gradually shifted from IaaS layer to PaaS layer.Kubernetes is a representative project of the new generation of PaaS.How to improve the utilization rate of resources is still a problem of resource management in cloud platforms.This paper first introduces the architecture and design of Kubernetes.Then,based on solving the problems of lag and low resource utilization in resource scheduling of cloud platforms,this paper studies the models that can be used to predict the workload of cloud platforms.A hybrid forecasting model based on ARIMA model and LSTM neural network model is proposed as a load forecasting scheme of cloud platforms.Then,combined with the prediction results,Kubernetes resource scheduling module and elastic scaling module are optimized.Finally,through a series of experiments,it is verified that the resource scheduling scheme of Kubernetes in this paper can effectively improve the resource utilization and reliability.