Resources @NERSC

The National Energy Research Scientific Computing Center (NERSC) is a high-performance computing center based in Berkeley, CA, USA.

Getting an account

Anyone may apply for a NERSC account. You can follow these instructions to obtain an account. You will need to fill in this form. Please specify m5197 as the project (PI: Zach Marshall). It is useful if in the description you include a short note about what muon collider projects you'll be focusing on. After a vetting period (this might be only a few hours, or two weeks for citizens / affiliates of specific countries), the account managers will be able to approve your account.

You will be asked once a year, around new years, to confirm that you would like to retain your account for the following year.

Available resources

The main system at NERSC today is Perlmutter. For the muon collider community, in allocation year 2026 (which roughly aligns with calendar year 2026), we have 10,000 CPU node-hours and 1,000 GPU node-hours. Batch jobs are charged against this quota, depending on the queue that you use for your jobs. Please try to use shared resources when possible to reduce the charge for jobs. Time on login nodes does not count against this quota. The time of an interactive batch job does.

Each user is given 50 GB of disk space in their home area. There is also a common disk area at /global/cfs/cdirs/m5197 with a 100 TB quota. Generally speaking, you should use that area for data storage, particularly when you expect to share data with others.

Each user account is generally set with a maximum number of CPU and GPU hours, so that no one accidentally consumes the entire annual allocation. This maximum can easily be increased by one of the account managers. The account managers will also monitor the storage and usage of CPU and GPU, and request more in case the group is running low (generally, we expect this to be possible).

Archival (tape) storage is also available via the HPSS system.

Getting help

There is quite extensive documentation at NERSC describing how to run jobs, best practices, where to store input data, and so on. Please refer to that documentation for any questions first. Lots of things are possible, including making web-viewable sites, sharing data, accessing cvmfs, using containers, using pre-installed scientific software, and so on; all that and more is documented on the site.

If the documentation is insufficient for any reason, you can try the NERSC channel in the muon collider Slack, Mattermost, or email. For general NERSC issues (e.g. password resets) you can try this documentation. If it seems like NERSC might be down, you can check the site status page, and you can subscribe to the outage calendar if you wish.