CPU and Memory Usage

After submitting a job, you can connect to the compute node where your job is running via SSH:

CODE

ssh username@compute_node

Once logged in, you can monitor your job’s CPU and memory usage using the following commands:

Lists all processes owned by the user.

CODE

ps -u username -o %cpu,rss,args

This shows instantaneous usage each time you run the command.

These commands run interactively and display live CPU and memory usage statistics.

CODE

htop -u username

You can use arrow keys or function keys inside htop to sort and filter processes by CPU, memory, or time usage.

For more details, refer to:

Displays GPU information including driver version,*
*GPU utilization, and GPU memory usage.

CODE

nvidia-smi

Example Output:

The example above

CODE

15MiB / 16376MiB

shows the amount of GPU memory used versus total available GPU memory.

Consider whether your job requires that much available GPU memory, as efficient resource usage improves cluster performance for all users.