Summit Transition Announcement and Overview

Fri, 17 Feb 2017 -0700

All CU-Boulder researchers are invited to move computational work to the new Summit supercomputer.

Here is an overview of the steps needed to begin running jobs on Summit.

Get a Research Computing account

If you already have an RC account, for instance if you are already using Janus, Blanca, or the PetaLibrary, then that account can also be used to access Summit.  

If you need to request an RC account you may do so at .  You'll also need to set up a 2-factor authentication method as described at .

Log in, check available software, and compile any needed applications

SSH to as usual.

Software modules for Summit are available in the Lmod hierarchical module system.  If you are not already using these "new" modules you can switch to them per .  On Summit, you will see newer versions of many modules than are available on Janus or the login nodes.  From a login node, SSH to a Summit compile node ("ssh scompile") to access Summit-specific modules.


Many software packages are already installed as modules, but if you need to compile your own application, please follow the guidelines at .

Understand Summit's storage systems

Review paying particular attention to the section on "Summit scratch".

Request an allocation of CPU time

Janus allocations do not transfer to Summit.  Please email to request access to a Summit "general" allocation.  Include 2-3 sentences describing your proposed usage. The "general" allocation can be used to ensure that your application works as expected, to check its optimization, and to understand how it scales as it runs on more CPU cores or nodes. This information will be needed when writing an eventual project allocation request.

Modify your job scripts or workflow for Summit

The Slurm batch/queue system for Summit is slightly different than on Janus. contains updated information about how to submit jobs to Summit.