Inria computing resources
There are several options to start using GPU resources for AI computing (we focus here in GPU resources because to the date, these are the most demanded/popular kind of resources used in AI).
Here is a brief description of the “general purpose moyens de calcul” (intranet link) at Inria.
Local servers
If you are in a Inria’s team, you can probably acces some machines for exclusive use of the team (we call them “local servers”). There exists some desktop servers equipped with general use GPU cards like Nvidia GTX/RTX cards, or even more powerful machines as Nvidia’s DGX gamme.
Local platforms GPU friendly
There also exists local platforms with GPU resources. These plateforms are managed locally, but you can ask for an access or login direclty with your Inria’s id, even if your team is not localized at the enter that manage the platform.
PlatFRIM (Bordeaux)
IGRIDA (Rennes)
Cluster NEF (Sophia Antipolis)
Cluster CLEPS (Paris)
Comparaison
Platform | No GPUs | Launcher | File system | Interface - Storage | Access |
---|---|---|---|---|---|
PlatFRIM | 82 | slurm | BeeGFS/Lustre | 10Gb/s | on demand |
Igrida | 52 | OAR | NFS | 10Gb/s / Infiniband - CephFS | Inria’s LDAP |
NEF | 152 | OAR | BeeGFS | Infiniband - SSD | on demand |
CLEPS | 16 | slurm | Lustre | Infiniband | Inria’s LDAP |