+a set of natively installed recipes for their users [@noauthor_undated-bt], or for researchers to develop and deploy their own software via containers. Even well-known package managers like Spack [@noauthor_undated-ae] and EasyBuild [@noauthor_undated-dj] expose software as modules. However, these package manager approaches don't always ensure reproducibility, or ease of development for the researcher. They typically require relying on some subset of system software, the underlying operating system, or even making changes to the system, which is not under the researcher's control. Although using containers in this context has been discussed previously [@noauthor_undated-rj; @noauthor_undated-rc], the majority of these approaches and tools do not make the process of developing and installing container modules easy. The single researcher must either convince a cluster administrator to install dependencies needed for their software, or build a container and manually move and interact with it on the cluster. All of these small challenges come together to make it harder for a researcher to develop and manage their own software, and subsequently to share their approach to reproduce the work. Using Singularity, Podman, or other container technologies installed via Singularity Registry HPC offers a solution to this challenge. The only requirement is the container technology software, and writing a simple configuration file for the registry. By clearly defining commands, and pinning exact versions of scientific software, researchers on high performance computing
0 commit comments