HPC Cluster update

The purpose of this call for tenders is to provide the ILL with a new compute queue for its HPC cluster. We aim for a queue of 12 Supermicro compute nodes with at least 16 cores and one FDR Infiniband (56 Gb/s) interface. The HPC Cluster consists today of about 100 compute nodes organized in 5 queues, interconnected by a shared, ZFS-based storage accessed over Infiniband. Curently the HPC is operated by the solution « torque/PBS » and we aim a migration to « SLURM ». Requirements Each node should meet the following minimal configuration: - 16 cores, We would appreciate Intel processors for compatibility reasons - Hyperthreading deactivated - 96 GB RAM - 1 port : 1 Gb eth - 1 port Infiniband: FDR 56Gb - 1 system/OS disk One cable FDR for each node: one half each 1m and 2m