CCR is now supporting the use of containers within our high performance cluster environment. Container computing is most widely associated with Docker; however, there are other products available that provide similar capabilities. CCR has implemented Singularity, which allows users the ability to use various types of containers (including Docker containers) but without requiring administrative privileges on the servers. Though the use of containers has been very useful for some applications, it was not easy to port scientific and HPC-centric applications into containerized solutions. Singularity was developed to support this community and the unique requirements it has.
The necessary components have been installed on CCR's clusters so that users can begin using Singularity. There is an enormous volume of documentation online for Singularity and, rather than re-writing it, we will include links here to what we believe most users will need to get started.
Singularity Interface Overview
Building a Container from Singularity Hub or Docker Hub
Running NVIDIA GPU containers using Singularity
Running NVIDIA containers on the CCR cluster - example
Note: Users will have to first download/install singularity and build containers on their own computers. They can then copy the containers to CCR clusters when needed.
To learn how to install Singularity on your OS, please visit this page.
Syslabs also offers a remote build tool so you do not have to install Singularity on your own device.
Want to learn more about Singularity? You can watch one of our workshop videos on the topic here.
An Intro to Singularity lesson is available as part of The Carpentries Incubator. This is a good lesson for learning the basics of Singularity. Most, if not all, of this should be possible to run on CCR's systems. However, this is not our work product so if you run into any issues or questions, please refer back to the content provider.