Biologists and neuroscientists at Stanford have standardized on Singularity for deploying and executing their BIDS (Brain Imaging Data Structure) stack and bridging their development environments (desktops, workstations, and local servers) to traditional HPC. The workflows can now be distributed, archived and leveraged as well as run on any HPC resources that have Singularity installed. Stanford (among other sites) is now running these Singularity containers in production on their local HPC system (Sherlock) as well as on XSEDE resources like Comet at the San Diego Supercomputing Center and Stampede at the Texas Advanced Computing Center.
Developed by the IT Division HPC Services architect Gregory Kurtzer, Singularity facilitates application and workflow portability across different environments including HPC by allowing the user to package their applications, libraries, and operating environment into an image or container that can be run elsewhere. In HPC environments, Singularity containers have access to local and shared filesystems as well as having support for MPI. Users who develop their applications on their laptop or desktop now have a way to run their personal environment and applications on a HPC cluster.
Next week, there will be a presentation on Singularity at the IT Division’s LabTech 2016 event on Tuesday Oct 11 and also at the portable-hpc.net DOE Lab conference on Weds Oct 12.