Design And Implementation Of The Auto Scaling Of The Cross-host Container Cluster Based On Docker

In a typical application scenario, the server would often be faced with unpredictable load demand. You need to add more servers to cope with the extra load when the application starts to get more traffic. When the traffic of the application begins to reduce, you need to terminate underutilized servers. To respond to such application scenes, cloud platform provides auto scaling service. Auto scaling, the key service of a cloud platform, automatically achieves the expansion or shrink of IT resources according to a variety of user-defined policies (schedules, operational status, etc.). The service can maintain sufficient performance while avoiding over-provisioning of IT resources, while reducing part of the management fee.Docker is an application container engine based on Linux Containers(LXC). It belongs to the operating-system-level virtualization, mainly used to solve the issues of how to rapidly build, deploy and share applications, and it can achieve lightweight virtualization. Currently, the auto scaling of cloud platform is mainly based on the traditional virtual machine technology (VM), which has a greater overhead on the resource utilization and start-up speed and performance.By studying the use of Docker as the underlying virtualization technology of cloud platform, this paper takes full advantage of the rapid deployment and efficient virtualization benefits of Docker, implements the auto scaling system of cross-host container cluster based on Docker, which greatly improves the resource utilization of cloud platform, reduces unnecessary waste of resources, is able to respond more quickly and effectively to unpredictable load demand, at the same time provides high throughput and low latency of quality service.The main research work of this paper includes:(1) The two key issues in the deployment of Docker inter host container cluster are network communication and file sharing. The vSwitch Open virtual switch is used to realize the communication between the host computer and the GRE tunnel, which is based on the combination of local image warehouse and NFS. (2) Design and implement a Docker based scalable group, Haproxy as the front load balancing, Etcd task configuration database, Confd dynamically generated Haproxy configuration script to achieve no sense of service refresh, improve the application of fault tolerance and availability. (3) The design and implementation of the Docker cluster resource monitoring module, calculate the real-time resource usage of the expansion group, provide the trigger mechanism based on the load of the automatic. (4) Using Docker as the base of the container cluster as the bottom platform, and combining with the Docker container resource monitoring and Docker, the design and implementation of a load based automatic telescopic system, and the system is tested to verify the feasibility and effectiveness of the system.