The HPC now supports containers! Using a new tool called Singularity, you can import and run Docker containers. This allows you to take better control of your runtime environment and provide more reliable workflows.
Software containers have drastically changed the way in which developers and systems adminstrators develop and deploy software. The technology has come to the world of HPC via a new platform called Singularity.
Packaging up HPC jobs and applications in Singularity containers offers HPC users several benefits:
- Control - You no longer need to rely on RCC staff to install specific software packages for you. Using Docker, you can now build your own runtime environments (images), upload them to the HPC as Singularity images, and run your jobs in them.
- Up-to-date software - If you need a specific version of a software package that is not installed on the HPC, you can install that version in your container
- Consistency - Applications and environments that are packaged as containers are insultaed to software upgrades, version changes, operating system
- Portability - Once you create a Singularity container, you can run it on the FSU HPC, the Open Science Grid, your own cluster or desktop, or any other environment that supports Singularity.
- Pre-packaged workflows - There are already thousands of images available on Singularity Hub available for download and immediate use. This saves researchers the time and effort of setting up your own containers for common scientific software packages.
To get started, check out our documentation:
We believe that containerized software is the way forward for scientific computing on HPCs, and we will continue to develop our support for containers. We invite your feedback on our documentation and our support for containerized worfklows.