CalcUA infrastructure

The CalcUA infrastructure currently consists of the cluster Hopper, providing a total of 168 compute nodes which deliver more than 75 teraflop of computing power.

Our old cluster Turing was put out of commission in January 2017, and a new cluster will be installed starting February 2017.

Cluster hardware

Hopper

  • 168 compute nodes (HP SL230s Gen8) for a total of 1632 cores:
    • 144 nodes with a dual Intel ten-core Xeon E5-2680 v2, 2.8 GHz, 64 GB Ram, FDR10 IB
    • 24 nodes with a dual Intel ten-core Xeon E5-2680 v2, 2.8 GHz, 256 GB Ram, FDR10 IB
  • 4 login nodes (HP DL380p Gen8)
  • 2 master nodes (HP DL380p Gen8)
  • 4 storage nodes (HP DL380p Gen8)
  • 100 TB central GPFS storage (DDN SFA7700)

Description of the infrastructure (March 2016).

Cluster software

User software

  • Applications : ABINIT, CP2K, Gaussian, Gromacs, NWChem, Quantum Espresso, R, Siesta, ...
  • Compilers : J2EE, GCC, Intel Cluster Studio
  • Libraries : Intel MKL, OpenBLAS, FFTW, HDF5, OpenMPI, ...

System software

  • Operating system : CentOS 7.X
  • Scheduling subsystems : Torque and MOAB
  • Monitoring subsystems : Ganglia, Nagios and CMU