Idiomatic and Reproducible Software Builds using Containers for Reliable Computing

Abstract: Containers as the unit of application delivery are the 'next big thing' in the software development world. They enable developers to create an executable image containing an application bundled with all its dependencies which a user can run inside a controlled environment with virtualized resources. Complex workflows for business-critical applications and research environments require a high degree of reproducibility which can be accomplished using uniquely identified images as units of computation. It will be shown in this thesis that the most widely used approaches to create an image from pre-existing software or from source code lack the ability to provide idiomaticity in their use of the technology as well as proper reproducibility safe-guards. In the first part, existing approaches are formalized and discussed and a new approach is introduced. The approaches are then evaluated using a suite of three different examples. This thesis provides a framework for formalizing operations involving a layered file system, containers and images, and a novel approach to the creation of images using utility containers and layer donning fulfilling the idiomaticity and reproducibility criteria.