FAQ

WAVE HPC (High Performance Computing) is Santa Clara University’s shared high-performance computing cluster used by students, faculty, and staff for computational research and coursework. It provides access to:

• Login nodes

• Compute nodes (CPU & GPU)

• High-memory resources

• Storage arrays

• Slurm job scheduling

• Open OnDemand access

Because it is a shared community resource, all users must schedule jobs through Slurm for fairness.

The WAVE support team can be contacted via:

wavesupport@scu.edu

How do I submit an account request?

From the documentation flow:

1. Submit an HPC account request form.

2. Your PI (Principal Investigator) must approve it. Usually your PI would be a faculty member or lead researcher who is responsible for a research project and oversees the use of resources like WAVE HPC.After approval, your account is provisioned.

3. You will receive login instructions.

If your PI does not receive the approval email, they should:

• Check spam

• Confirm correct email was entered

• Contact wavesupport@scu.edu

From the SSH setup guide:

Step 1 – Generate Key

Open terminal (Mac/Linux/WSL):

ssh-keygen -t rsa -b 4096 -C "username@scu.edu"

• Save to default: ~/.ssh/id_rsa

• Optional: set passphrase

This creates:

• Private key → id_rsa (keep secret)

• Public key → id_rsa.pub (send to WAVE)

Step 2 – Copy Public Key

Automatic method:

ssh-copy-id username@login.wave.scu.edu

Manual method:

cat ~/.ssh/id_rsa.pub

Copy output → paste into:

~/.ssh/authorized_keys

Then fix permissions:

chmod 600 ~/.ssh/authorized_keys

Step 3 – Test Connection

ssh username@login.wave.scu.edu

If successful → no password prompt.

WAVE uses modules to manage software environments.

Modules allow users to:

• Load different software versions

• Avoid conflicts

• Customize environment

Common commands:

module avail
module load <software>
module list

Users can:

1. Load Python module

2. Use Conda environments

Example:

module load python
conda create -n myenv python=3.x
conda activate myenv

• R is available via module system.

• RStudio Server may be accessible through Open OnDemand.

• Allows browser-based R sessions on compute nodes.

What is Slurm?

Slurm (Simple Linux Utility for Resource Management) is the job scheduler used by WAVE HPC to:

• Allocate compute resources fairly

• Manage queues

• Run interactive and batch jobs

It is used by ~60% of top supercomputers.

Login Node vs Compute Nodes

When you log into WAVE:

You first access a login node.

Login node:

• Configure environment

• Write code

• Compile

• Test small datasets

Note: Large jobs running on login nodes will be killed.

Compute nodes:

• Used for heavy computation

• Accessed via Slurm

How do I submit a Slurm job?

Create a job script:

#!/bin/bash
#SBATCH --job-name=test
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=4
#SBATCH --mem=32G
#SBATCH --time=04:00:00
./my_program

Submit with:

sbatch job.slurm

Add GPU request:

#SBATCH --gres=gpu:1

Or in interactive mode:

srun --gres=gpu:1 ...

Here's the edited source code: ```html

From the documentation:

srun --partition=cmp --nodes=1 --ntasks=1 --cpus-per-task=4 --mem=32G --time=0-04:00:00 --pty /bin/bash

or:

srun -p cmp -N 1 -n 1 -c 4 --mem 32G -t 0-4 --pty /bin/bash

What this does:

• Allocates a compute node

• Opens interactive bash session

• Waits if resources unavailable

When session closes then resources released.

squeue -u username

Check detailed info:

scontrol show job <jobid>

sacct

For usage statistics.

sinfo

Cluster status is also visible via:

Open OnDemand → WAVE Cluster Status

Shows:

• Jobs running

• Jobs queued

• CPU usage

• GPU usage

• Node availability

Only the following are backed up nightly:

• /WAVE/users

• /WAVE/projects

Other directories are NOT backed up.

Do not store irreplaceable data outside backed-up directories.

Causes:

• Public key not in ~/.ssh/authorized_keys

• Wrong permissions

• SSH agent not loaded

Fix:

chmod 600 ~/.ssh/authorized_keys
ssh-add ~/.ssh/id_rsa

1. Generate new key.

2. Send new public key to wavesupport@scu.edu

3. Or manually update authorized_keys.

Remove old keys after rotation.

They will be terminated without warning.

Periodically for security:

• Generate new key

• Update authorized_keys

• Remove old keys