Research And Implementation Of Image Combination Based On Docker

In recent years,the development of cloud computing is unstoppable.Using cloud computing technology,users can significantly reduce the interaction with the service provider.Do not need too much management work,but users can real-time,on-demand from the resource sharing pool to obtain the necessary resources services.Docker appeared in early 2013,now has a pivotal position in the field of cloud computing.Docker takes "container" technology as a foundation and core,and use namespace,cgroups and other technologies for the container to provide the resource isolation and security.Compared with traditional virtual machine technology,Docker is more lightweight,more rapid start-up speed.Typically,a physical machine can run hundreds of containers.In addition,Docker technology has a cross-platform,only to build an image,then the host is equipped with a Docker can run the corresponding container.Based on Dockerfile,this paper puts forward the markup language DIML(Docker Image Markup Language)which is used to make the Docker image rapidly.Users can use DIML to make different components into the corresponding image template,and then select images as required to generate a multi-component composite image.DIML implements the reuse of component templates and the syntax optimization of Dockerfile,reducing the number of image layers,reducing the size and speeding up the creating of image.By modifying the Docker source code,the paper improves the way it uses the image to generate container.When the native Docker is building the container,it will mount all the layers of the image,plus a layer of read-write.After modification,the user can specify the image part of the layer and the read-write layer for union mounting,on-demand to generate the container operating environment.Finally,the paper studies how to build Hadoop cluster on Kuberntes platform.Compared with the traditional way to build Hadoop clusters,this way is more convenient and faster.And Kubernetes as a distributed architecture,to protect the Hadoop cluster stability and scalability.