Provisioning of docker containers with TOSCA

Abstract

In order to master the administration and automation problem of distributed applications in the cloud age, topology & orchestration platforms have been established in the past few years. Application topologies and their entire lifecycles can easily be modeled and later on be deployed on various cloud environments. Standards like the Topology and Orchestration Specification for Cloud Applications (TOSCA) help to keep the description of applications platform independent and increase interoperability between components. Another recent paradigm in Cloud Computing is containerized virtualization. The particular and significant popularity of Docker containers was mainly driven by the needs of having less dependencies when moving from development to production environments. The technology around Docker container still evolves very fast and projects to provision and manage Docker container in a automated way have already been adopted by major Cloud providers (e.g. Amazon ECS, Azure Container Service, Google Container Engine), but lack in topology & orchestration platforms like Cloudify or OpenTOSCA. The cloud provider offerings use container cluster technologies like Apache Mesos or kubernetes under the hood, as the lifecycle management of containers is a complicated task. Container cluster technologies provide an easy way to automatically scale, deploy and manage multiple Docker container on various infrastructures. This thesis aims to enable the support for the deployment of clustered Docker containers using a TOSCA compliant topology & orchestration language and execution environment. More specifically, the Cloudify environment is used as the basis to enable the modeling and deployment of container clusters hosted on kubernetes. By the usage of the Cloudify platform the interoperability with other non-containerized applications and general platform independency is assured, while still taking advantage the container cluster features. The resulting system is able to orchestrate, manage and scale application components individually, regardless of the underlying cloud technology.