Research And Implementation Of Container Elastic Scaling System Based On Microservice Architecture

With the continuous development and change of cloud computing related technologies,the traditional cloud computing model with virtualization technology as the core has undergone significant changes.As a representative of container technology,Docker has set off a wave of a new generation of resource computing mode.The existing elastic scaling methods are difficult to cope with the large-scale dynamic change of the cloud environment due to the limitation of the lack of flexibility of the physical machine or the limitation of the timeliness of the resource supply of the virtual machine.Container engine technology represented by Docker container effectively has solved these problems.Aiming at the problems that the existing elastic scaling strategies have single elastic scaling index and the unreasonable elastic expansion causes the waste of resources,a container elastic scaling method based on machine learning XGBoost prediction algorithm is proposed.In this method,the resource index and performance index of the container are used as the elastic scaling characteristic data,and the number of elastic scaling containers in the next time window is predicted by using the pre-trained XGBoost prediction model.The Kubernetes container orchestration tool is then used to assist with container expansion or shrinkage operations.Compared with the existing elastic scaling methods,the elastic scaling method proposed in this thesis has the lowest response time and the lowest error rate in the dynamic cloud environment,and the cloud service default rate is less than 0.5%.Based on the proposed elastic scaling method,a container elastic scaling system based on micro-service architecture is designed.Firstly,the overall hierarchical architecture of the container elastic scaling system is designed with the design concept of micro-service architecture.The system is divided into four parts,including the front-end display layer,the service management layer,the core business layer and the data storage layer.Secondly,the functional modules of the core business layer are designed in detail according to the business requirements,which mainly include four modules: system management module,cluster data monitoring module,elastic expansion module and cluster management module.Finally,the basic components of microservice and the core modules of the system are designed and programmed,and the core modules of the system are tested by the black box method.The test results verify the effectiveness of the proposed method.