Lawrence Berkeley National Laboratory, United States of America
Reproducibility in computational workloads is a critical factor in scientific and commercial applications but it is often elusive. There are many factors that contribute to this challenge. In this paper, we discuss some of those sources and describe how containers can play a role in addressing this challenge. The role of workflow tools, best practices and gaps is also presented. The goal of this paper is to provide readers with a basic understanding of the issues and equip them with practices to use containers to improve the reproducibility of their work.