Research On Container Elastic Scaling Technology Based On Load Prediction

With the continuous development and improvement of container technology,container clouds represented by Docker and Kubernetes have received extensive attention and applications in the industry.Although Kubernetes has powerful functions and advanced design,there are still some problems that cannot be ignored.The most prominent problem is that the existing elastic scaling mechanism of Kubernetes is too simple and cannot be well adapted to various application scenarios in the production environment.Therefore,in order to solve such problems,this article proposes corresponding improvement schemes,which can effectively improve the existing elastic scaling function of Kubernetes.In order to obtain the load changes of the applications deployed on Kubernetes in advance,a cloud resource load combination prediction model VMD-TCN that combines variational modal decomposition(VMD)and temporal convolutional network(TCN)is proposed.According to the nonlinear and non-stationary characteristics of the cloud resource load sequence,this model first uses VMD to decompose the original load sequence into different frequency and relatively stable modal components,and then uses TCN to predict the decomposed components,thereby improving The prediction accuracy of the model is improved.At the same time,in order to enable Kubernetes to use the prediction results of the VMD-TCN model to guide the elastic scaling process,a load prediction-based resource elastic scaling strategy LPB-HPA(Load Predicting Based Horizon Pod Autoscaling)is proposed.This strategy comprehensively considers the relationship between the predicted load value and the current load value and the resource threshold.According to the different characteristics of the expansion phase and the shrinking phase,the appropriate algorithm is used to calculate the expected number of copies of the application,and then the calculation is completed according to the results.The expansion or contraction action avoids affecting the quality of service of the application due to delayed expansion or frequent expansion and contraction problems.Experiments show that the VMD-TCN combined prediction model in this paper has achieved better prediction results than other comparison models,and improved prediction accuracy.At the same time,the LPB-HPA proposed in this paper can execute better scaling decisions based on the prediction results of the VMD-TCN model,reduce the occurrence of untimely expansion and frequent scaling,and achieve the purpose of reducing application response time and improving service quality.