HPC wiki

Information about the HPC clusters of the University of Twente.

For an new account on any of the clusters refer to the Cluster table below to see who is responsible for account creation on the cluster.

All home directory's are mounted on /home after a users in logged in.
Software is located under /software.
You can set the path and other variables to use the software by using the module for the software.

There is a “scratch” space on all nodes in /local.
You can create a directory for yourself and put some temporary files there.
This can benefit the speed of your job because the local storage is much faster. Your home directory is on NFS which is slow.
This “scratch” space is automatically cleaned every day if the files are not changed for 90 days!
If you just copy or move the files from this space to your home directory as the last command in your job you won't loose anything.

- Connect to the cluster using SSH
- Transfer files to the cluster
- Create a job script and submit/schedule your job

• Do you need a module
• do you need to copy data to and from the node

- Wait until your job gets queued, it gets executed and it finishes
- Study the results generated by your job

• remove data from the cluster afterwards to free diskspace

To access the cluster you will need to be working on a PC connected to the UT network or use EduVPN.

Refer to the Cluster table below to see which headnode to use.

OpenSource SSH clients:
- PuTTy; SSH CLI terminal.
- Windows Terminal; Terminal for Windows systems only.
- MobaXterm; Enchanced terminal for windows with X11 server, tabbed SSH client.

Refer to the connecting page for more information.

See Environment Modules on SourceForge for more in depth information.

The clusters have quite a few programs pre-installed. They are managed as optional modules that you have to load before you can use them. One reason for using them this way is that this makes it possible to have multiple versions of the same program on the same system. Below there is an basic explanation of module usage. - Using modules
module av to list all available modules on the cluster.

------------------------------------------------ /software/Modulefiles -------------------------------------------------
bison/2.7.1   centaur/14    clang/3.5.0          matlab/R2012b  ofed/3.18                    paraview/5.8.0
castep/20.11  centaur/14.1  clang/3.6.0          matlab/R2017a  openfoam/20.06
centaur/11    centaur/14.5  cuda/7.0             matlab/R2020b  paraview/5.4.1
centaur/12    centaur/15    firedrake/firedrake  metis/5.1.0    paraview/5.6.1
centaur/13.5  centaur/15.1  mathematica/11.2     miniconda/3    paraview/5.6.1-GUI(default)

module load <module name> to load a available module.
module list to list all currently active modules.

Currently Loaded Modulefiles:
 1) matlab/R2020b   2) paraview/5.6.1

module unload <module name> to unload a loaded module.
module purge to unload all loaded modules.

Refer to the Cluster table below to see the cluster software details.

Refer to the corresponding scheduler page for more information :

• Slurm or the Slurm quickstart
• Hadoop/Yarn.
• t.b.d.

We use 2 different types of accounts: AD or local.

If you login with you AD account you can change you password by typing: passwd
This can have a big impact because this account is also used for all of the other services on the University. Think twice before you use this.

If you use a local (NIS) account you can change your password by typing: yppasswd