A new HPC modular infrastructure
This new facility as been built in 2021.
A new concept of modular datacentre infrastructure for high performance computing
More rapidly deployed and extensible than a classical building, a joint CEA-ATOS project, it is comprised of 72 containers prepared at the factory. It hosts a ATOS XH2000 supercomputer, pre-assembled at ATOS factory, with blades inserted in the containerised racks on premise.
A more rapidly deployed solution
- Standardised modules
- Prepared at the factory
- Assembled on premise
A generic building block, or module, comprises of 4 prefabricated elements, 2 for IT service and 2 for technical service, for a power capacity of 2.5 MW max.
A ‘slice’ (see below) is an independant half-module, whose failure does not impact the rest of the system, not even the other slide in the same module. Self-contained local technical equipment in slices makes extra power and cooling redundancy unnecessary.
Compute building block = module
- 1 module = 2 slices = 4 containers ( 2 IT containers + 2 technical equipment containers)
- 2.5 MW (2MW for IT)
- Slice = ½ independant half module
- No power supply nor cooling N+1 redundancy
- Incident impact limited to a ½ module