Languages:   Polski  • English
Okeanos foto.jpg
Instalacja: Cray XC40
Nazwa: Okeanos
Typ procesora: Intel Xeon E5-2690 v3
Architektura: x86_64
Reprezencja danych: little-endian
2.6 GHz
Liczba procesorów
w węźle:
2 x 12 rdzenie (HT2)
Ilość pamięci
w węźle:
128 GB
System plików: Lustre (rozproszony system plików)
System operacyjny: Linux
System kolejkowy: Slurm
Wszystkie komputery | okeanos | boreasz | nostromo | hydra | topola
Podręcznik | Programowanie

Spis treści

System architecture

The Okeanos system is a Cray XC40 supercomputer. It is composed of 1084 computing nodes. Each of the node has 24 Intel Xeon CPU cores (x86_64 architecture [1], code name Haswell) with a 2-way Hyper Threading (HT) [2]. Nodes are connected with a Cray Aries Network [3], with a Dragonfly topology

Intel Xeon implements little-endian byte order[4]. Each node has 128 Gb of main memory (RAM). Memory is organized in ccNUMA (cache coherent Non-Uniform Memory Access [5]) domains.

Computing nodes do not have a local storage. All data is storaged on the parallel file system (Lustre).

Using the system

Okeanos is best suited for large-scale computations, which require several thousand cores and several ten terabytes of memory in single runs. Supporting such large-scale computations is Okeanos’s primary purpose.

The system is available in the frame of type A, B and C Research Computational Grants.

Access to the system

Access to the Okeanos system is available with secure shell (ssh). User sessions are started on a login-nodes, visible as okeanosX(where X = 1,2,3), using round-robin order. Login-nodes are only for jobs preparation, data transmission and program compilation. It is not allowed to run any computation or load intensive applications on the login-nodes. User jobs are submitted to the batch system and executed on computing nodes.

Note: login-nodes CPU architecture (Intel Sandy Bridge) and computing nodes (Intel Haswell) are different; users should take notice when compiling programs.

Batch system and user jobs

The system is controlled by Slurm resource manager (batch system). Jobs should be submitted with either salloc or sbatch Slurm commands.

To prepare computing job, user should create dedicated script with resource specification and commands to execute application. Example scripts are available on page about Slurm/en.

File system and data storage

The system is using parallel file system for all data storage. File system implementation is Lustre [6].


No limits on user job are defined at the moment. Single job allocation is limited to 256 computing nodes.

User environment

The system is running Linux operating system (specific version of the SLES distribution).

User environment (programming, applications) is set up with modules tool. For instruction on setting environment use command: man module; basic usage:

module avail for a list of available modules

module list for a list of currently loaded module

module load module_name to load a given module name

module unload module_name to unload a given module name

Available applications

Scientific applications are installed in /apps directory tree.

  • abinit
    • abinit/8.0.8
    • abinit/8.4.2
    • abinit/8.4.4

  • ansys
    • ansys/17.0
    • ansys/17.1
    • ansys/19.1

  • barkeleygw
    • barkeleygw/1.2.0_c
    • barkeleygw/1.2.0_r

  • cp2k
    • cp2k/2.6.2
    • cp2k/3.0

  • elk
    • elk/4.0.15
    • elk/4.3.6

  • espresso
    • espresso/5.2.1
    • espresso/5.3.0
    • espresso/5.3.0_slow
    • espresso/5.3.0_up
    • espresso/6.1
    • espresso/6.2.1

  • fluent
    • fluent/17.0
    • fluent/17.1
    • fluent/18.0
    • fluent/19.1

  • gromacs
    • gromacs/4.5.7
    • gromacs/5.0.4_plumed-2.1.1
    • gromacs/5.1.2
    • gromacs/5.1.4
    • gromacs/5.1.4-plumed2
    • gromacs/5.1.4_intel
    • gromacs/2016.1
    • gromacs/2016.4

  • lammps
    • lammps/30Jul16
    • lammps/31Mar17

  • namd
    • namd/2.11
    • namd/2.11_memopt

  • ncbi-blast
    • ncbi-blast/2.2.28
    • ncbi-blast/2.2.31
    • ncbi-blast/2.3.0

  • python
    • python/anaconda2.7
    • python/anaconda3.6

  • vasp
    • vasp/4.6.1
    • vasp/5.2.12
    • vasp/5.4.1
    • vasp/5.4.1_p2
    • vasp/5.4.1_p3
    • vasp/5.4.4
    • vasp/5.4.4_optics


  1. x86_64 - processor architecture
  2. Hyper Threading - hardware multi-threading implementation
  3. Cray XC Network - Cray Dragonfly topology
  4. little-endian - byte representation
  5. Non-uniform memory access
  6. Lustre - parallel file system