highlights

We describe challenges in reproducing complex high energy physics (HEP) applications.
We propose an invariant framework for conducting reproducible computational science.
Light-weight virtualization approaches can help capture the dependencies of an application into a self-contained package and reproduce the application through the package.
Standard software delivery mechanisms can help distribute the preserved packages.
We report the performance overhead of running, packaging, and reproducing two HEP applications through the framework.