At the ARI we have a small cluster of 12 linux machines with slightly different configurations (some use GPUs, some don't, some have more memory then others, ...). Originally used for computing complex BEM problems, it grew over the years, and is now also used for creating, testing, and experimenting with new models, setups and tools in the area of machine learning. Having this cluster offers us the possibility to:

  • quickly install, test and compare different software packages (e.g. octave with different toolboxes or different versions of libraries like blas, atlas or the intel mkl) or different versions and configurations of these packages,
  • do some calculations that are too big for local personal computers
  • have easy access to a linux environment with open programs like octave, R, c and fortran compilers, cuda, python or several open numerical libraries, and many more.

Configuration

  • Most of the machines are equipped with 64GB RAM, if more memory is needed two machines are equipped with 128GB.
  • The majority of the machines are equipped with Intel Core i7 processors with 8 Cores
  • 4 of the machines are additionally equipped with an NVIDIA graphics card designed for GPU computing
  • The home directories of the users are located on a shared external mass storage (25TB NAS)
  • The cluster devices are connected via a local 10GB LAN and with 1GB to the LAN of the institute