First steps
The following instructions are intended for new users of our cluster. It explains some basic terms and contains links to further information. It explains in detail what a cluster is, how it differs from a normal computer, how you can access the cluster and where you can go if you have any questions.
What is a cluster?
A cluster is first and foremost a large computer that is made up of many smaller computers. A cluster is divided into so-called nodes, each of which has its own working memory and a certain number of processors, also known as cores. The OMNI cluster, for example, has around 450 compute nodes on which calculations run, 4 login nodes that are directly accessible to users and a number of specialized nodes. The normal compute nodes have 64 CPU cores and 256 GB of RAM, a more detailed list can be found here. The individual nodes in the OMNI cluster do not have their own hard disks, but share a central file system.
ZIMT is currently still operating the HoRUS cluster, which will be replaced by the OMNI cluster after a transition phase.
Linux
Like practically all clusters, our cluster is operated with Linux. If you are not (yet) familiar with Linux, we offer several ways to familiarize yourself with it.
Firstly, we offer an introductory Linux course every semester, alternating between German and English. Our Linux course usually takes place towards the end of the lecture period. You can find an overview of our course dates here.
Secondly, the ZIMT has created the HPC-Wikiin cooperation with other universities in NRW.
in cooperation with other universities in NRW. On the HPC-Wiki you will find a video tutorial on Linux, which is similar in content to our Linux introductory course.
Finally, on this pagewe describe
some basic Linux terms.
Differences between a cluster and a normal computer
Many things about a cluster are identical to any other (Linux) computer. However, there are also differences. The most important difference is that you do not sit in front of the cluster itself, but connect to it from another computer. This is shown in the following illustration:
As you can see, you always connect to one of the login nodes. You usually only interact with the compute nodes indirectly via the SLURM scheduler, which will be described later. In addition, all nodes share the file system, i.e. the hard disks of the cluster.
Here are some other important differences:
Workspaces
You have a home directory on the cluster as with all Linux systems. However, if you are performing calculations, you should create a so-called workspace. Workspaces are located on a different part of the file system, which has a faster connection to the compute nodes. In addition, unlike the home directory, they are unlimited in size. However, they expire after a while and are then deleted. Details on workspaces can be found here.
The OMNI cluster has a so-called burst buffer. This is a partition of SSDs (solid-state drives) that are intended for jobs that write or read a particularly large number of files. We also describe the use of the burst buffer on this page.
Modules
Operating systems use so-called environments to decide which program is executed with a particular command. For example, the file paths to the executable files and various other settings are stored in environment variables. As many users with different requirements work on a cluster, different versions of the same software often need to be available and the commands would be ambiguous if everyone used the same environment. This is why the environment on the cluster is modular and can be easily exchanged.
For many programs installed on the cluster, environments are predefined in so-called modules. You can simply load the required modules. How to do this is described here
described here.
How do I access the cluster?
The cluster is available to all employees of the University of Siegen free of charge and they can also register students, who can then also use the cluster. All you have to do is activate your user account for the cluster and then set up a connection from your computer to the cluster.
Account activation for university members and students is described here
described here.
The connection is established using the Secure Shell Protocol (SSH) and is generally possible from Windows, Linux and Mac OS. The setup of an SSH connection is described here
described here.
You will be informed of the address of the clusters by email when your account is activated. The clusters are accessible from the university network (including university VPN) and, with restrictions, also from the internet.
How do I bill on the cluster?
Calculations on the cluster are carried out in so-called jobs. You define for a job how many resources (CPUs, memory, etc.) you need and for how long. You also describe, usually in a so-called job script, which programs are to be executed during the job. You then place the job in a queue and the SLURM scheduler decides when to run it based on the job size and other factors. You can find out how to create job scripts and set up jobs here, more information about Slurm can be found here.
Help with questions
The cluster website covers the most important topics and refers to the corresponding documentation for installed programs. If you are looking for help with a specific Linux command, the built-in help functions will often help you. Under Linux, this is the manual function in particular, for which you enter
If the information on this website does not help you, you can send an e-mail to hpc-support@uni-siegen.de or come to our office hours
office hours. These are also your best points of contact for problems with the cluster or installation requests.
ZIMT also offers a training program on all aspects of high-performance computing and cluster use. Upcoming training courses can be found here.
ZIMT also offers support in the development and optimization of your software. ZIMT experts are available to go through your software with you or to advise you personally in other ways. If you would like advice, you can also send an e-mail to hpc-support@uni-siegen.de.