Research And Implementation Of Elastic Cloud Platform Based On Docker

Nowadays,the Internet is growing at an alarming rate,and Web applications have permeated every aspect of people's life.Cloud computing provides a flexible and powerful computing model that allows developers to access resources on-demand and pay as they use.Cloud computing attracts an increasing number of developers to migrate their Web applications to cloud platforms.The load of Web application usually varies drastically along with time.Cloud platforms should provide elasticity ability to change the amount of resources allocated to a Web application in order to meet the actual varying demands because of the changing workload.Traditionally,elastic clouds are built on virtual machines(VMs),VM is too heavy for Web applications and VM itself wastes a lot of resources.In this paper,we first analyze and define the problems Web applications are facing.Secondly,we compare Docker container and virtual machine and study current researches on elastic cloud.Thirdly,we design and implement DoCloud which is an elastic cloud platform for Web applications based on Docker containers instead of virtual machines.In DoCloud,we adopt horizontal scalability growing or shrinking the number of container instances to change a Web application's resources and we build a hybrid elasticity controller that incorporates proactive model and reactive model for scale out coupled with proactive model for scale in.For developers' convenience,we integrate a private Docker registry into DoCloud,developers can ship and deploy Web applications more easily.DoCloud also supports fast hot-upgrade to achieve high availability of Web applications.At last,we evaluate the efficiency and scalability of DoCloud.We use Tsung to simulate three types of load in different scenarios.Our experiments show that DoCloud can dynamically adjust resources of applications within seconds according to the varying load and maintain higher resource utilization in a single container.DoCloud can help developers save much cost and help cloud providers improve the efficiency of the resources utilization on real hardware hosts.We also evaluate the hot-upgrade function,the result shows that developers can upgrade their Web applications without any downtime.