Singularity paper published by SRCC team member Vanessa Sochat
SRCC team member Vanessa Sochat recently published a paper on Singularity, a HPC containerization technology that we (and many others) are very excited about. Singularity enables two things that are a big deal: user defined containers that run safely on a shared compute cluster and reproducibility of scientific computing workflows. Many dependencies that would normally have to be installed on the cluster can now be installed by a users in their Singularity container that they build on their local desktop or laptop.
These containers can then be run on a HPC cluster via a scheduler like SLURM as a standard compute job. Complex workflows can also be encapsulated and shared with Singularity containers thus greatly aiding scientific reproducibility. Several users on our Sherlock cluster have run Bioinformatics and Machine Learning workflows using Singularity, thus skipping many of the software and driver dependencies that come with running cutting edge research code on a large cluster. Singularity Hub is a container registry for which Vanessa is lead developer, and aims to offer builders and storage for Singularity containers. The online interface provides users with an easy way to connect to Github, and automatically build images in the cloud.
