Abstract

In the Virtualization industry, container virtualization has increased in popularity with the emergence of container engines in 2013. Container engines create, remove and manage containers. They do so by utilizing container images from which containers are created. Images work as templates, are reusable and can be uploaded for others to download. Big contributors in the field of container virtualization grouped up and founded the Open Container Initiative. The project established an industry standard for the architecture of container images and how they should be run.

The IT consultant company Zenon has requested a thesis to be done in the field of container virtualization as they are thinking of starting to work with one but want to learn more about modern solutions. Due to the lack of back-to-back comparisons of modern solutions supporting the standard set by the Open Container Initiative, this thesis makes an evaluation of available systems in industry standard terms of performance. The company is also interested in fields; isolation, image management and time to build an image of a Zenon application and deploy a container from it.

The evaluation in this thesis was made on container engines Docker and rkt. Results showed that Docker performed better in all benchmarks. Rkt starts a container in less time than Docker but takes longer to build the specified image. Docker has much more capabilities in terms of image management with the ability to create images from scratch and to run on multiple operating systems, while rkt has limited ability for creating images, requires more configurations and is exclusive available for Linux. Both container engines offer varied isolation, where Docker has 2 security profiles and rkt lets the user implement custom isolation profiles with several provided options available.