Research On Prediction And Configuration Of Container Cloud Resource Based On LSTM

Containers have driven the rapid development of container clouds by virtue of their flexibility,efficienct and fast.Application containerization has become the first choice for individual or enterprise users.However,due to the short development time of container technology and low maturity,in large-scale cluster applications,complex resource management problems are often faced,and with the development and application of container technology,the development concept of DevOps is deeply rooted in people's minds.The demand is also constantly changing.The demand for high concurrency,high availability,high flexibility,and high flexibility is getting stronger and stronger.How to make the resource rational and efficient configuration under the premise of ensuring the safe and stable operation of the cloud computing environment becomes One of the hot issues in current field research.This thesis mainly studies the resource prediction and configuration of the container cloud platform from the following three parts:(1)A container cloud resource prediction model based on improved genetic algorithm for optimizing LSTM neural network.Resource prediction is the analysis of the historical data to extract the potential characteristics of the data before and after the correlation,and then model it to predict the resource load data of the resource in the future.In this thesis,we use the global optimization ability of genetic algorithm and improve the genetic algorithm.Combine LSTM NN to model the resource load time series data,realize the high resource prediction accuracy rate through intelligent parameter adjustment.(2)A new container cloud platform structure design and implementation proposal.For container-oriented applications,it has the absolute advantage of resolving applications and dependencies.With hardware-assisted virtualization technology,it achieves system-level security isolation and high flexibility for hardware resources,and solves the security and isolation problem caused by shared kernels in containers.Virtualize physical hardware devices through KVM,establish a shared virtual resource pool,and build a CaaS service platform using Docker and Kubernetes.(3)A container cloud cluster resource allocation mechanism based on predictive model.As the container provides strong encapsulation and isolation for the application,there are more and more container instances in the container cloud cluster,and the resource allocation research requirements in cluster units are becoming more and more urgent.Based on the constructed resource prediction model and the defined resource prediction feedback mechanism,this thesis proposes a predictive-based container cloud resource adaptive elastic configuration strategy and mechanism to allocate resources for existing or newly added task requests.The resource prediction,resource monitoring,and resource allocation are closely linked to achieve rational utilization and configuration of the container cloud platform cluster resources.Finally,simulation and comparison experiments are carried out on existing physical equipment and public datasets.The experimental results show that the improved genetic algorithm is more effective in the optimal solution combination search of LSTM NN structure parameters,and the prediction model has higher accuracy.The container cloud platform design scheme is highly feasible,and the resource allocation strategy and mechanism improve the utilization of cluster resources.