Partenaires

Université Paris-Est
Ecole des Ponts ParisTech
IFSTTAR
CNRS


Rechercher


Accueil du site > Equipes de recherche > Multi-échelle > Multi-scale computing resources

Multi-scale computing resources

This page gives an overview of doubs, the HPC server in use at the Multi-scale team. Requests for installation of additional software or account creation should be addressed to S. Brisard.

Contents

Technical specifications

  • HP Proliant DL580 Gen8
  • CPU : 4x Xeon E7-4860 v2 (2.6 GHz, 12 cores, 24 threads)
  • RAM : 1024 GB (DDR3 SDRAM, 1866 MHz)
  • Storage : 6TB (RAID-5)
  • System : Ubuntu 16.04LTS

Account creation

Requests for account creation should be addressed to S. Brisard. Each user can store data in two different places

  • /home/dupont : personal home directory,
  • /libre/dupont : additional space to hold large chunks of data.

It should be noted that quotas are set up for the personal home directories, which are limited to ***TODO : check the value ***. Therefore, users are strongly invited to store their data in the /libre partition.

RAID-5 configuration of the hard drives offers some redundancy that should avoid losses of data should a crash occur. However, the machine is not backed up on a regular basis : long term storage remains the responsibility of the users. Likewise, users are asked to remove from their directories data that are no longer used.

Python

Prepackaged versions of Python 2.7 and 3.4 are installed system-wide. The REPL is invoked through the commands python and python3, respectively.

However, in order to benefit from the latest version of all libraries in the SciPy stack without breaking dependencies of the Ubuntu packaging system, virtual environments have been set up using virtualenv.

It is therefore recommended to invoke Python through these virtual environments. To activate the desired virtual environment, the command source /opt/py27/bin/activate must be issued in a bash console for Python 2.7. For Python 3.4, the command is source /opt/py34/bin/activate. The prompt should change to something like

(py34)dupont@doubs : $

Regardless of the Python version, the REPL is then invoked through the unique command python. The whole SciPy stack is installed in both virtual environments. The command deactivate deactivates the virtual environment currently in use.

FreeFem++

An MPI-enabled version of FreeFem++ is installed in /opt/freefem++-3.49-1. The binaries can be found in /opt/freefem++-3.49-1/bin.

Code_Aster

Version 12.5 of Code_Aster is installed in /opt/aster. The binaries can be found in /opt/aster/bin.