Research And Application Of PaaS Platform Based On Docker

The concept of cloud computing covers the abstract integration,dynamic flow and efficient use of resources.Cloud computing aims at abstracting the resources from the physical environment for more flexibility.In the cloud platform,users could manage,monitor and schedule resources through a unified interface.This platform also integrate the loose functions and provides a large number of services.The traditional cloud platform uses the virtual machine as the basic scheduling unit.However,the PaaS platform in this paper uses the Docker container instead of the virtual machine.Docker is a container technology based on LXC(Linux Container).Compared virtual machines,the container has the much smaller size of resources.From the flexibility point of view,deployment time of the container is shorter and the container is more portable.Therefore,using Docker to improve the cloud platform would be more suitable in the future.This thesis used Docker container as the underlying virtualization technology to design a PaaS platform.It also analyzed the PaaS related technology and carried on with the cluster overall design and proposed the concrete design plan,including Mesos resources scheduling,Docker network,NFS storage,the container resources limit and so on.The requirements of the platform have been researched and each function module has also been designed and implemented.The platform uses the container as the application node and Spark to process the request.Implemented an application release platform,which uses the Docker container as a unified development environment.It uses SVN to manage code,Jenkins to complete the integration and construction and docker-compose and Marathon to complete the application of the layout and deployment.The platform also has a high availability based on load balancing and flexible scalability.Load balancing is through the Nginx agent,but the flexible expansion strategy is designed independently.Using CAdvisor to monitor the resources and provide a unified view.While designing the log system and the user management system of the platform.As the elastic algorithm is improved,the system has been optimized.Finally,the system is deployed and tested.The test includes elastic test and application release test and the results of the test are analyzed to verify the availability of the platform.This platform has improved the basis of the traditional PaaS platform based on the advantages of the Docker container.In the meantime,it provides a new way of thinking for the design of a cloud platform.