CEA pre-production computing center for R&D
The machines at the OCRE computing center are a test bed for new hardware and a collaboration around new/experimental hardware that covers a large number of technologies in all areas of HPC :
OCRE offers an HPC innovation space shared by users and system administrators who can develop and evaluate scientific applications and administration tools in an R&D environment as close as possible to production environments.
The facility has a training and collaboration space, a working environment for trainees, and the teams offer training on HPC topics.
OCRE is a computing center accessible from the internet.
One of the OCRE supercomputers in Bruyères-le-Châtel. © CEA
The heterogeneous computing cluster includes several partitions covering a large number of technologies :
- Various CPU and GPU architectures
- Multiple types of interconnection networks
- Around ten types of nodes, 400 calculation nodes, more than 12,000 CPU cores
Cluster based on the Open Source “slurm” scheduler.
Tools for developers :
- Virtualization, containers with pcocc
- Continuous integration with non-regression testing tools (gitlab, jenkins)
Access to new developments or innovative services.