Dark mode Light mode

Getting Started with Great Lakes

Content

• Available Resources
• Creating an ARC account
• Configuration
  • Recommended Configuration for CPU Workloads
  • Recommended Configuration for GPU Workloads
  • Other Configurations
• Launching a Session
  • Launching an Interactive Session
  • Launching an SSH Session
  • Creating a Job Template
• Useful Resources

---

Important Notes

• You must connect to U-M VPN to access Great Lakes and Lighthouse if you’re not connected to MWireless.
• We can only use up to five GPUs as a whole at the same time. Please terminate your session when you finish your work so that others can access the GPUs.
• Delete your job if you finish before the time requested ends. You will only be charged for the time when the machine is up and running.
• The availability depends on the resource and the amount of time requested. If the wait time is too long, try reducing the requested time or switching to another partition.
• All the GPUs on Great Lakes are fast enough for most cases.

---

Creating an ARC account

1. Request an HPC login here

   Please put “Accessing Great Lakes for research projects” in the request.

   

2. Log into your account in the web portal

   You must connect to U-M VPN to access Great Lakes if you’re not connected to MWireless.

   

---

Configuration

Recommended Configuration for CPU-only Workloads

• Partition: standard
• Number of cores: 4 / 8 / 16 / 32
• Memory (GB): 20 / 40 / 80 / 160

Click to show the screenshot

Recommended Configuration for GPU Workloads

There are three GPU-available partitions. spgpu is usually the one with the least wait time, which has 224 NVIDIA A40 GPUs (48GB VRAM). On spgpu, you can request 4 CPUs and 48 GB RAM without extra cost for each GPU requested on the spgpu partition, so the following is the most cost-effective option.

• Partition: spgpu
• Number of cores: 4
• Memory (GB): 48
• Number of GPUs: 1

Click to show the screenshot

Other Configurations

For each GPU requested, Great Lakes provides some CPU cores and RAM memory without additional charge. Please use the following configurations for each partition to maximize the freely-provided CPUs and memory (as of Winter 2025; see the current configurations):

Partition	CPU cores	Memory (RAM)	GPU	GPU speed	GPU memory	#GPUs available
spgpu	4	48 GB	A40	faster	48 GB	224
gpu_mig40	8	124 GB	A100	fastest	40 GB	16
gpu	20	90 GB	V100	fast	16 GB	52
standard*	1	7 GB	-	-	-	-

Here is the cost for each partition (as of Winter 2025; see the current rates):

Partition	Hourly cost	CPU hours equivalent
spgpu	0.11	7.33
gpu_mig40	0.16	10.66
gpu	0.16	10.66
standard*	0.015	1

---

Storage

• Home directory
  • 80 GB limit per user
  • Available at /home/{UNIQNAME}
• Scratch space
  • 10TB limit as a whole
  • Data may be removed if not used in 30 days
  • Available at /scratch/pat463f25_class_root/pat463f25_class/{UNIQNAME}

---

Launching a Session

Launching an Interactive Session

1. Navigate to “Interactive Apps” on the web portal for Great Lakes or Lighthouse

2. Select the application (here we take JupyterLab for example)

   

3. Specify the configuration of the instance (see below for suggestions)

   Click to show an example configuration for a CPU only workload on Great Lakes

   

4. Connect to the launched instance

   

5. Make sure to delete the job if your workload is finished earlier

Launching an SSH Session

This will create a persistent compute node that won’t end automatically. Please make sure to kill the job after your workload is finished so that the resources are released.

1. Navigate to “Job Composer” on the web portal for Great Lakes or Lighthouse

   

2. Click “New Job > From Default Template”

   

3. Click “Open Editor”

   

4. Replace everything in the editor with the following and click “Save”

   This defaults to requesting 1 GPU with 8 CPUs and 62 GB RAM on Lighthouse. See below for configuration suggestions.

   #!/bin/bash
   #SBATCH --account=aimusic_project
   #SBATCH --partition=aimusic_project
   #SBATCH --cpus-per-task=8
   #SBATCH --mem=62G
   #SBATCH --gres=gpu:1
   echo $SLURMD_NODENAME
   tail -f /dev/null
   

   

   Click to show an example configuration for GPU workloads on Great Lakes

   #!/bin/bash
   #SBATCH --account=hwdong0
   #SBATCH --partition=spgpu
   #SBATCH --cpus-per-task=4
   #SBATCH --mem=48G
   #SBATCH --gres=gpu:1
   echo $SLURMD_NODENAME
   tail -f /dev/null
   

5. In the “Jobs” tab, select the created job and click “Submit”

   

6. At the right panel, click the “slurm-{ID}.out” file

   

7. Copy the node name

   

8. To access the compute node, you will need to first SSH into the login node

   ssh {UNIQNAME}@lighthouse.arc-ts.umich.edu
   

   And then you can SSH into the compute node

   ssh {NODENAME}.arc-ts.umich.edu
   

9. Make sure to delete the job after your workload is finished

   

Creating a Job Template

You can create a template to save some time from editing the job script every time.

1. Navigate to the “Templates” tab

2. Click “New Template”

   

3. Enter the template name and click “Save”

   

4. Select the created template and click “View Files”

   

5. Click the triple dots next to “main_job.sh” and click “Edit”

   

6. Replace everything in the editor with the following and click “Save”

   This defaults to requesting 1 GPU with 8 CPUs and 62 GB RAM on Lighthouse. See below for configuration suggestions.

   #!/bin/bash
   #SBATCH --account=aimusic_project
   #SBATCH --partition=aimusic_project
   #SBATCH --cpus-per-task=8
   #SBATCH --mem=62G
   #SBATCH --gres=gpu:1
   echo $SLURMD_NODENAME
   tail -f /dev/null
   

   

   Click to show an example configuration for GPU workloads on Great Lakes

   #!/bin/bash
   #SBATCH --account=hwdong0
   #SBATCH --partition=spgpu
   #SBATCH --cpus-per-task=4
   #SBATCH --mem=48G
   #SBATCH --gres=gpu:1
   echo $SLURMD_NODENAME
   tail -f /dev/null
   

7. Return to the previous page and click “Create New Job”

   

---

Useful Resources

• Guide to Using Great Lakes
• Great Lakes Cheat Sheet
• Great Lakes User Guide

---