resources-available.markdown

layout
default

Resources available

Compute nodes

The compute nodes are the physical machines where you can run Virtual Machines (VMs).

• 21 HPC compute nodes, each node having:

  • 64 vCPU cores
  • 256 GB RAM
  • 3.2 TB local SSD disk
  • 2x 10 Gbit network connection per compute node

• 10 HPC compute nodes, each node having:

  • 80 vCPU cores
  • 512 GB RAM
  • 3.5 TB local SSD disk
  • 2x 10Gbit network connection per compute node

• 7 HPC compute nodes, each node having:

  • 80 vCPU cores
  • 576 GB RAM
  • 3.0 TB local SSD disk
  • 2x 10Gbit network connection per compute node

• 6 GPU compute nodes, each node having:

  • 16 CPU cores
  • 4 GPU's of type NVidia K2 GRID, accessible through PCI passthrough, enabling applications to get the performance boost of the direct access to a GPU card
  • 256 GB RAM
  • 800 GB local SSD disk
  • 2x 10 Gbit network connection per GPU node

• 2 GPU compute nodes, each node having:

  • 16 CPU cores
  • 4 GPU's of type NVidia Tesla P100, accessible through PCI passthrough, enabling applications to get the performance boost of the direct access to a GPU card
  • 128 GB RAM
  • 800 GB local SSD disk
  • 2x 10 Gbit network connection per GPU node

Storage resources

• 510 TB Ceph distributed object storage with SSD caching
  • used in x3 redundancy
  • 10 Gbit network connection per storage node
  • Ceph storage is accessible as a block device within a VM