SIS has setup, is operating and supporting the powerful and reliable HPC clusters Euler (internal link) and Leonhard (internal link) as versatile resources for scientific computations at ETH Zürich.

HPC Cluster Euler.

Software Environment

SIS offers and supports a rich set of commercial and open source applications covering many research fields, for example:

  • Bioinformatics and life sciences:
    Bioconductor, BLAST, Bowtie, Cellprofiler, CLC Genomics Server, CuffLinks, FSL, TopHat
  • Computational fluid dynamics:
  • Finite element methods:
    Ansys, Abaqus, MSC Marc, MSC Nastran
  • Multi-physics phenomena:
    Ansoft Maxwell, COMSOL Multiphysics, Trilinos
  • Quantum chemistry and molecular dynamics:
    Abinit, ADF, CP2K, GAMESS, Gaussian, GROMACS, LAMMPS, Molpro, NAMD, NWChem, Orca, Q-Chem, Quantum ESPRESSO, Turbomole, VASP
  • Symbolic, numerical and statistical mathematics:
    Maple, Mathematica, MATLAB, Octave, R, Stata
  • MATLAB Distributed Computing Server (MDCS) for offloading computationally-intensive calculations from your workstation to a computer cluster as transparently as possible
  • Visualization:
    FFmpeg, GDL, IDL, Splash, VisIt

For most of these applications, we keep several versions side-by-side to ensure the continuity of long-term research projects spanning many years. In addition to these centrally installed applications, experienced users have the possibility to develop and run their own software on the cluster. If necessary, SIS will be happy to assist with software installation and optimization. We provide a variety of development tools for the most common programming languages (C/C++, Fortran, Java, Python):

  • Compiler:
    GCC, Intel, LLVM, Open64, PGI
  • Scientific libraries:
  • MPI libraries:
    MVAPICH2, Open MPI
  • Build systems:
    Apache Ant, CMake, GNU autotools, make, qmake
  • Version control:
    CVS, Git, Mercurial, SVN

(Internal link to a list of third-party applications.)

History of central HPC Clusters

ETH IT Services has been a pioneer in cluster computing since the end of the 1990's. Its first large Beowulf-type cluster, Asgard (1999-2007), contained over 500 processors and was then the largest in Europe. Its successors, Hreidar (2004-2008) and Gonzales (2005-2008) formed the basis of a new platform for cluster computing called Brutus. According to the June 2009 Top-500 list, Brutus was at that time the most power-efficient general purpose computer in the world. These machines were built and operated in the division ID Systemdienste which is in charge of server computing and storage. When Scientific IT Services were formed in 2013, Brutus and the team supporting it moved to the new division.

The Brutus cluster has been continuously upgraded and expanded over the years. It currently contains nearly 19,000 processor cores, which give it a theoretical performance of 195 teraflops (1 teraflop = 10^12 floating-point operations per second). Unfortunately this growth could not continue indefinitely due to lack of space, power and cooling on the site of ETH in Zurich. The IT Services thus decided in 2013 to build a second large cluster in ETH's brand-new computer room at CSCS in Lugano. This new cluster, called Euler, was installed in 2014 and has been expanded several times already. The cluster is constantly extended and upgraded. For the current state, have a look at the scicomp wiki (internal link).