Link Search Menu Expand Document
Bernardi Group Resources Sharing
Table of contents
  1. Apply for an Account
  2. Organization
  3. Job Submission
  4. Others Useful Stuff
  5. Notes for the Project Lead

Apply for an Account

To apply for an account, your Principal Investigator should already have an ongoing project and a certain number of allocated hours on NERSC. To create an account, follow the steps given in the below link: apply for an account.

The project ID for Bernardi research group is m2626. Once the relevant information has been submitted, the account will be approved by PI (Prof. Bernardi), and you can start using NERSC. For more information on setting up Multi-Factor Authentication, managing your account, tracking your usage, etc. , please refer to the NERSC documentation.

Organization

When we log into Cori by ssh <usrname>@cori.nersc.gov, we are at our home directories where we only have a few memory size (e.g., a few tens of GB) so we cannot run large calculations there.

To run and store our calculation results, we can change into a scratch directory (https://docs.nersc.gov/filesystems/cori-scratch/) /global/cscratch1/sd/<usrname> where we can create many directories for large calculations. However, the data that are not used for 12 week in the scratch directory will be purged (https://docs.nersc.gov/filesystems/quotas/).

Please back up your data in the scratch directory regularly.

Job Submission

A small working example

NERSC uses the slurm queue system. A sample example of a job script (job_script.sl) is shown below:

#!/bin/bash
#SBATCH -N 20                # Number of nodes 
#SBATCH --tasks-per-node=32  # Number of processors per node
#SBATCH -C haswell           # Machine on which you want to submit your job
#SBATCH -q regular           # Queue in which you want to submit your job
#SBATCH -J Job_Name          # Job Name. The same name will be displayed in the queue
#SBATCH -t 00:30:00          # Maximum walltime for your job.

#OpenMP settings:            # Open MP commands to enable threading within the processors
export OMP_NUM_THREADS=2     # Number of threads per processor

#run the application:
srun -n 640 -c 2 --cpu_bind=cores myapp.x - Run the file using the appropriate executable and the command srun

Once the resources required for the job have been specified, it can be submitted to the queue via the command sbatch job_script.sl. The job status can be viewed using commands sqs and squeue. Job can be deleted mid session using scancel job_id.

Generate a job script

If you want to generate a job script on NERSC, please use the following link: job generator. The website is very straightforward to use, but please modify the example script to match your application.

More details

please refer to the NERSC documentation.

Others Useful Stuff

  • To get the first hand information for user information, please consider joining the NERSC user Slack
  • To know the live status of Cori: motd

Notes for the Project Lead

Every year we need to submit a request to renew our NERSC allocation. Please start to prepare the relevant proposal or document more than 1.5 month before the submission deadline. For more details, please see NERSC allocation request form. You can find the related files for our previous requests in our shared Caltech Box.