Supplement for "Container-driven Reproducible Research with Development Containers"

The following repository contains a number of example Docker based Development Containers from our paper. All containers are listed under the .devcontainer directory with an additional SLURM script each can be tested on the following systems:

• Devcontainers 1,2,3,5: Any system using Docker container engine
• Devcontainer 4: Any system with a CUDA 12 enabled Nvidia GPU
• slurm-gpu.sh: TACC Stampede3 cluster

Usage

Before using this, you must have Visual Studio Code (VS Code) and the Dev Container extension downloaded and installed. Additionally, the latest version of Docker Desktop or Docker Engine is required to run this on a local machine. For more information, check Appendix A in the paper for installation instructions. GitHub Codespaces is an optional alternative and can be launched via the "Code" dropdown menu in this repository:

Note: Multiple container profiles used in this repository are not supported by GitHub Codespaces and therefore are only compatible with VS Code. There is a default container that follows ./.devcontainer/3-jupyter schema that will launch if using Codespaces.

1. Clone the repository on your desired system:

git clone [email protected]:dddlab/container-computing.git

2. Open the cloned folder using VS Code.

3. Using the "Remote Window" interface, select "Reopen in Container":

4. Select the desired container to build:

5. If you wish to run the SLURM script, simply schedule it to be run by using sbatch slurm-gpu.sh on a compatible system.