Research On Flexible Scaling Strategy Of Container Resources Based On Kubernetes

Cloud computing is a service mode based on the network to provide computing resources for users on demand.Docker container technology has been widely used in cloud computing environment with the advantages of small resource consumption and fast start-up.In the container cloud environment,a large number of containers need to be managed efficiently and conveniently.In many container arrangement systems,kubernetes has been widely used because of its excellent design and good scalability.However,the native responsive elastic scaling strategy of Kubernetes has shortcomings.In the capacity expansion stage,there may be a problem of user request response delay;in the capacity reduction stage,deleting redundant pod copies may affect the cluster load balancing.In this paper,the following research work is carried out for the above problems:(1)The docker container technology,kubernetes cluster architecture and core concepts are introduced,and the elastic scaling strategy of kubernetes is studied.The existing problems and the causes of the problems are analyzed.This paper studies the basic principles of autoregressive integrated moving average model(ARIMA)and grey model,optimizes and improves the traditional grey GM(1,1)model,combines the ARIMA model with the improved grey GM(1,1)model,constructs the combined forecasting model,and forecasts the load change of the cluster.(2)Aiming at the problems of kubernetes in the expansion stage,the optimization scheme is proposed,and the predictive expansion strategy is designed and implemented.A combined forecasting model is established to predict the cluster load,and the capacity expansion operation is completed in advance according to the prediction results,so as to avoid the problem of user request response delay that may occur during the peak load.(3)In order to solve the problem that kubernetes may affect the load balance of the cluster after deleting the replica in the capacity reduction phase,in the capacity reduction phase,in order to solve the impact of the capacity reduction action on the resource balance of the cluster,based on the state priority,increase the calculation of the load balance weight,design and implement the optimal capacity reduction strategy combining the load balance factors.(4)A series of experiments are designed to verify the prediction effect of the load prediction model,and kubernetes cluster is built to verify the feasibility and effectiveness of the predictive expansion strategy and load balancing reduction strategy.In the experimental environment,when the load reaches the peak value,the response time of the predictive expansion strategy is 47% less than that of the original expansion strategy;after the reduction operation combined with the load balance optimization,the average resource utilization variance of each node in the kubernetes cluster decreases from 0.0807 to 0.0026.The experimental results show that predictive capacity expansion can effectively solve the problem of response delay in the capacity expansion stage and reduce the response time of user requests.The optimized capacity reduction strategy can optimize the cluster load balancing problem without affecting the quality of service..