aurorascience

  • Increase font size
  • Default font size
  • Decrease font size
AuroraScience
AuroraScience in brief

Aurora-Science is a research project at the crossroad of computational sciences and Computer Architecture.
It builds on the combined know-how collectively available to the members of the collaboration on:
  • design, development and operation of application-driven high-performance computer system (e.g., the series of APE machines, developed by INFN).

nodecardchassis

  • algorithm development and physics analysis in compute-based areas of physics (Lattice Gauge Theory, Computational Fluid-Dynamics, Molecular Dynamics, few- and many-body Nuclear Physics), quantitative-biology (protein folding), Bio-Informatics (Gene-Sequencing) and medical physics.
fig_temperaturefig_quark

 

Aurora-Science can be seen as a scientific project enabled by leading-edge computational systems and by specific competences in the useful operation of these systems. The main goals of the Aurora-Science Project are the following:

  • Tailor the architecture of a massively parallel computer system to the specific needs of a large class of regular computational problems. This is done by assembling a large number of latest generation multi-core CPUs (Intel Nehalem, to be soon replaced by Intel Wesmere) and interconnecting them with a low-latency 3-D toroidal grid. The project uses the hardware recently introduced by Eurotech in their Aurora class of machines and re-designs and optimize the (FPGA-based) network processors of the torus network.
  • Procure and operate a medium size Aurora prototype machine, optimized for Aurora-Science. The envisaged installation will have in excess of 100 processing nodes with a target peak performance of some tens of TeraFlops. The project considers the prototype as a convincing test proof that large scale systems can be efficiently operated by a combination of off-the-shelf hi-end processors and application-driven interconnects. Following successfull operation of the prototype, the project will consider enlarged collaboration options, with the main goal of developing a PFlops class machine optimized for scientific applications and fully available to basic science.
  • Modify, tune, port and optimize for the Aurora-Science prototype key computational algorithms in the scientific areas described above, and show effectiveness for state-of-the-art scientific simulations.
The structure of the project

AuroraScience is operated on the basis of a research agreement between IstitutoNazionale di Fisica Nucleare (INFN) and the Provincia Autonoma di Trento(PAT). INFN and PAT jointly provide the Aurora-Science budget.

August 1, 2009 and will continue till the end of January 2011. A further phase of the project (stretching over 18 more months) is forecast but not yet approved by INFN and PAT.


The project is managed by ECT* in Trento. The scientific team of the project includes members from:

  • ECT* - Project coordination, Lattice QCD and Nuclear Physics
  • INFN (Ferrara, Milano Bicocca, Parma) - Torus network architecture, Application-specific system software, LQCD, Fluid-Dynamics
  • Physics Department, University of Trento, Trento - Molecular Dynamics, NuclearPhysics
  • DEI (Department of Information Engineering, University of Padova) - Network architecture, application-independent algorithmic optimization
  • ATreP (Agenzia Provinciale per la Protonterapia) - Medical physics
  • FEM (Fondazione Edmund Mach, San Michele all' Adige) - Bio-informatics, gene sequencing.

A more in-depth description of the AuroraScience computer is available here

For further information contact Gigi Scorzato (scorzato (at) ect.it)