Research And Implementation Of Method For Prediction Of Container Resource Load

With the development of Docker container technology,more and more enterprises are beginning to use Docker containers to build cloud platforms.In order for an application hosted on a Docker container to run securely,it will be fully resourced.However,in most cases,the hosted application is not running at the highest load,and the resources such as CPU and memory are not at the highest load at the same time,so the pre-provisioned resources are idle most of the time.This has caused a waste of resources.When the application is in a high load state,the pre-allocated resources may not be sufficient.The specific prediction algorithm is used to predict the resource requirements of the container and the application,and the resource allocation optimization is performed in advance to improve the resource utilization and service.quality.Firstly,according to the characteristics of container environment and resource load dynamics,it is proposed to use ARIMA(Autoregressive Integrated Moving Average Model)and cubic exponential smooth combination model to predict the Docker container resource load by one-step prediction.The weighting values of the two models in the combined model are solved according to the sum of squares of their respective predicted errors for a period of time before the current prediction,and the weights vary with time series.According to the purpose and requirements of the combined model,a real-time prediction system for Docker container resource load is designed and implemented at the application level.The system implements collection,storage,prediction of container resource load data and dynamic scheduling of CPU and memory resource usage based on predicted values.The experimental results show that the combined model is better than the two single models in predicting multiple loads.Compared with the ARIMA model,the combined model prediction accuracy is improved by more than 10%;compared with the cubic exponential smoothing model,the prediction accuracy of the combined model is improved by more than 20%.