Virtual Desktop Infrastructure

University of Geneva provide virtual machines with GPU for image analyses or other application needing GPU.

These virtual machines are not intended to replace the use of the HPC clusters. The use is limited in duration.

These machines are not persistent. Make sure to save all you data to a network location when you are not using the VM anymore. All data saved on the VM are deleted after you log off or if you reboot it.

If you are disconnect from the VM, the VM will be kept for a certain amount of time (check the VM specifications). After this delay the VM will be deleted and it is impossible to retrieve any data.

Software preferences are saved across sessions only on Windows ?

At this time the virtual machines are free of charge. This might change in the future.

The service is currently in beta test.
You can request an access by sending an filling the support-si form

This is the procedure to access the VM :

1. Plone - Install VMware horizon client
2. Connect to the allocated pool using horizon view

The VM are ephemeral. It means that as soon you disconnect from the VMware horizon client, the VM is trashed.

If you reboot the VDI, it is thrashed as well!

The Linux VM are accessible through ssh for everyone who has an ISIS account at Unige. It's your duty to restrict the access to your user only.

Windows 10 Virtual machine :

• 8 Cores @2.30GHz
• 64 Go Ram
• 4 Go GPU ram on shared Nvidia P40 Grapic card
• 250 Go of scratch space on SSD (deleted after logoff)
• 10 Gbps access to UNIGE Nas (NASAC)
• Virtual machines are reseted every night at 4:00 am
• Installed softwares in images

Ubuntu 18.04 Virtual machine:

• CPU
• RAM
• 4 Go GPU RAM on shared Nvidia P40 Grapic card
• 250 Go of scratch space on SSD (deleted after logoff)
• 10 Gbps access to UNIGE Nas (NASAC)
• This is a special VM made as test with a very long session duration.
• The members of the AD group gl_vdi_gpu-linux-1 have sudo rights.
• Installed softwares in images

References:

https://hpc-community.unige.ch/t/new-service-on-demand-virtual-server-with-gpu/841