Administrative data
System namePhysOn
A brief description of the systemHigh-performance computing cluster
Technical specifications
Number of servers96
Server SpecificationSuperMicro CPU-Quad
Processor specificationXeon 5335; Xeon 5420; Xeon E5-2620; Xeon E5-2650v4; Tesla M2090
CPU cores on the server8, 8, 12, 24, 16
Server memory12GB, 16GB, 16GB, 64GB, 94GB
Total number of CPU cores216
Maximum number of parallel processes per processor240
Connection typeInfiniband DDR & FDR
Connection delay1.1 to 1.5 μs
Connection tape20 – 56 Gbps
Local file system typeNFS & Lustre
General data storage48TB
Type of acceleratorsnVidia Tesla M2090 & K80
Number of cores6200
Server Accelerators2
Servers equipped with accelerators1
Max. performance (Tflops, double precision)3.5
Real performance (Tflops, double precision)3.2
OSUbuntu and CentOS
Version16.06 LTE ; 7x
Task management systemSun Grid Engine & Slurm
Development ToolsIntel Vtune, Eclipse (incl. Intel XE support, PyDev, Photran), Nsight Eclipse (CUDA support), Nvidia Visual Profiler, Intel Inspector, Jupyter Notebook, Jupyter Console, PyCharm, git
Compilers and librariesКомпилатори: Intel Compilers, PGI Compilers, GNU C/C++/Fortran Compilers, OpenJDK; Интерпретатори: Python 2.7.x (Intel), Python 2.7.x (GCC), Intel Python 3, Python 3.6.x (GCC), TCL 8.5 (GCC), Perl 5.16 (GCC), Julia 1.1; Библиотеки/модули/добавки: Intel MKL, Intel DAAL, OpenMPI (Intel), OpenMPI (GCC), MPICH2 (Intel), MPICH2(GCC), HDF5 (Intel), HDF5+MPI (Intel), HDF5(GCC), NetCDF (Intel), NetCDF(GCC), PyDAAL, NumPy, SciPy, H5py, PyTables, mpi4py, pandas, matplotlib, sympy, tensorflow-gpu, MDAnalysis, OpenVINO, CUDA+cuDNN (8.x,9.x,10.x)
Application softwareGROMACS, Quantum Espresso, ABINIT, CP2K, elk, WRF, Maple, Matlab, Mathematica, Siesta, Kshell, GAUSSIAN, LAMMPS, Tensorflow (with GPU support, on Intel Python 3), NAMD, CHARMM, PG-Storm