Skip to main content

FAQ

HPC: OMNI

Here you’ll find answers to the most important questions about the HPC. The FAQ provides a quick overview of key topics, procedures, and points of contact, and helps you find your way around the HPC. If you can’t find the answer to your question here, please feel free to contact our HPC team.

For specific inquiries
→ support@zimt.uni-siegen.de

 

For office hours & training sessions
→HPC: Contact

 

→HPC: Home

Access

HPC: OMNI

We describe the processhere: .

 

SSH comes preinstalled on most systems.

If it isn't, search for the OpenSSH client.

... > For fields you need to fill out, you can omit the < and > characters in your entry. 
 

There are different key types, such as ED25519. RSA is often used in tutorials but is no longer secure. 
 

1) Check for existing keys

  • Go to your user directory / home directory
  • Check if the hidden .ssh folder exists.
  • If it exists, check to see if there is a .pub file and a key file with the same name inside it. 

2) Generate an SSH key

  • Open PowerShell/Terminal/Command Prompt
  • Use the following command:

ssh-keygen -t ed25519 -C " dein Kommentarfeld>"

 

  • This command includes the following:
    • The comment field is optional, but it may be helpful to include your device name. 
    • A password is recommended but not required.
    • -t ed25519 specifies the key type.
    • If you don't have any keys, you can accept the default path.
    • If you have other keys, you can use /home/user/.ssh/othername
  • After that, the file will be in the .ssh folder.

3) Add to the HPC

Send an email to support@zimt.uni-siegen.de

 

4) To log in

You’ll then receive your login credentials via email

Private key

  • is stored on your computer
  • belongs only to you
  • is never shared
  • can be protected with an additional password

Public key

  • may be shared
  • is stored on the server
  • is essentially the information that allows the server to recognize your private key

 

No, all members of the University of Siegen (both students and staff) can use the cluster free of charge. You may be required to pay a share of the license fees for certain software, such as 

Abaqus; in such cases, a warning will be displayed when you load the 

module.

 

 

No, unfortunately, industry-funded projects are not permitted to use our computing resources.


 

Yes, the OMNI cluster has 10 nodes with a total of 24 NVIDIA Tesla V100 GPUs; read more here.


 

Invoices

Even though the process of writing job scripts and submitting jobs could be simplified or automated even further, we have decided not to support this in most cases. This is because you share the cluster’s limited resources with all other users, and we want to encourage you to develop a basic understanding of how the system works and to be mindful of your resource usage.

No. The cluster's capacity is limited, and the SLURM scheduler schedules jobs using a fair-share/backfilling algorithm that ensures the shortest wait time for everyone. It is our stated policy not to interfere with this system.

The prioritization algorithm used by SLURM has four main criteria:

  • Fair Share: If a user consumes a lot of resources (i.e., a lot of CPU time) in a short period, the priority of subsequent jobs is reduced. This criterion is by far the most heavily weighted.
  • Age: The longer a job has been in the queue, the higher its priority.
  • Job Size: Large jobs are given slight preference to compensate for their longer wait time.
  • Backfilling: After jobs are sorted according to the three criteria above, SLURM fills any gaps with smaller jobs, which are thus processed sooner.

You can view the priorities using the sprio command.

Not for regular computing jobs. In exceptional cases, we set up reservations for specific courses so that participants can run their jobs directly during class. Please 

contact us if you are interested. 

 

There could be many different reasons for this. First of all, you should check the SLURM log file for error messages (especially at the end of the log file). If your job was aborted with the message “not enough memory,” you may have requested too little RAM in the SLURM settings ( --mem option); for more information,

here.

If you cannot identify the error, please send the log file and the job script, along with a description of the problem, to support@zimt.uni-siegen.de

Software

On our homepage, you'll find an overview of the installed software and available compilers and interpreters. If you already have cluster access, you can use ` module avail ` to view the installed modules (Note: Some software may not be available to all university members or may require a fee; in this case, you will receive a warning when loading the module).

We decide on a case-by-case basis whether to install something. Please contact us.

We decide on a case-by-case basis whether to update software. To keep maintenance efforts manageable and to ensure compatibility and security, we install software only from specific sources (such as the operating system vendor’s software repository), where the latest versions are not always available.


 

Files

Daily snapshots of your home directory are created automatically. If you don't wait too long (no more than 30 days), you can restore the files yourself. Instructions on how to do this are 

describedhere.

 

During the first 10 days after your workspace expires, you can restore it yourself; instructions on how to do so are 

described here.