• DocumentCode
    3146675
  • Title

    A Performance Model of Direct Numerical Simulation for Analyzing Large-Scale Systems

  • Author

    Kerbyson, DarrenJ ; Barker, KevinJ

  • Author_Institution
    Performance & Archit. Lab., Pacific Northwest Nat. Lab., Richland, WA, USA
  • fYear
    2011
  • fDate
    16-20 May 2011
  • Firstpage
    1824
  • Lastpage
    1830
  • Abstract
    In this work we develop, validate and use a performance model of the combustion code DNS3D that uses Direct Numerical Simulation to model turbulence in fluid motion. This is an important application area and is expected to use large fractions of future large-scale systems. From a thorough analysis of the code, its key computational characteristics are coupled with the performance characteristics of a parallel system using an analytical performance model that is parameterized in terms of key application and system aspects. The performance model is validated on two current systems, a multi-core AMD Opteron system with an Infiniband interconnect, and an IBM Power5+ system with an HPS interconnect. High prediction accuracies are demonstrated on these two systems. The model can be used to explore changes in the performance characteristics of either the application or of the systems, and in particular to examine the processing possibilities for DNS3D on future systems. We illustrate here how the model can be used to explore the impact of changes in the characteristics of either the system or of the application.
  • Keywords
    large-scale systems; multiprocessing systems; multiprocessor interconnection networks; parallel processing; DNS3D combustion code; HPS interconnect; IBM Power5+ system; Infiniband interconnect; direct numerical simulation; fluid motion; large-scale system; multicore AMD Opteron system; parallel system; turbulence; Adaptation models; Analytical models; Computational modeling; Libraries; Numerical models; Numerical simulation; Program processors;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Parallel and Distributed Processing Workshops and Phd Forum (IPDPSW), 2011 IEEE International Symposium on
  • Conference_Location
    Shanghai
  • ISSN
    1530-2075
  • Print_ISBN
    978-1-61284-425-1
  • Electronic_ISBN
    1530-2075
  • Type

    conf

  • DOI
    10.1109/IPDPS.2011.341
  • Filename
    6009051