Access to the Vega HPC cluster is provided via Nordugrid ARC middleware and direct access is provided via Slurm.
The access itself is free with
SiGNET Certificate for Slovenian users and
IGTF Certificate for EuroHPC users. Slovenian users can find more information on how to acquire the certificate.
EuroHPC users can find more information on how to proceed.
Obtaining a user account for direct access via the Slurm system
To use clusters via the Slurm system, the user must gain access to the login node. A login node is a server that is directly connected to the system and it is part of the cluster, properly configured to communicate with the control daemon and to have command-line tools already installed. To access the cluster via the login node, you can request a user account and register at SLING support - email@example.com. The username applies to the SLING infrastructure and can be edited on the SSO identity management web interface fido.sling.si. The command line and SSH protocol are usually used for access.
Once the access is approved, an account will be created and credentials will be sent by e-mail, which includes a temporary password. This password is only temporary and does not enable login to the login nodes and needs to be changed on fido.sling.si. Vega allows login only with an SSH key, so your public key must be uploaded under your account on fido.sling.si. You can find more information at link.
Access via Nordugrid ARC middleware system
To access via the Nordugrid ARC system, the user must obtain a SiGNET/IGTF certificate. Once the certificate is downloaded, you have to upload it to the browser and visit the website voms.sling.si, enter the required information and confirm enrolment in a virtual organization. After filling out the form, you will receive an email where you need to confirm your email address. You just have to wait for the confirmation of membership in the virtual organization, which you also receive at your e-mail address. In order to submit jobs via ARC, users then just have to install and configure ARC Client on their system.
Submitted jobs must have a time-limited duration (you cannot use the system to host permanent services), but a single user can send large number of tasks which can be stored and later retrieved from other nodes and can run on multiple servers simultaneously. The system also supports multithreaded tasks:
OpenMP tasks (interprocessor communication in one node) and
MPI (multiprocessor task with communication between nodes over the low-bandwidth InfiniBand network).