STARTSEITEKontakt | Lokales | Impressum | Englisch
Minerva Logo of the MPG
Photo of the Institute
Banner Max-Planck-Institut
Fachgruppe Computational Methods in Systems and Control Theory

Diese Seite wird nicht mehr aktualisiert. Bitte besuchen Sie unsere neue Webpräsenz.

This page is not updated any longer. Please visit our new website.


Projekt aus MPG Mitteln

otto

Der Linux Cluster des MPI Magdeburg



Projektleiter:
  • Prof. Dr. Peter Benner
    Max Planck Institut for Dynamics Complex Technical Systems Magdeburg,
    Computational Methods in Systems and Control Theory,
    Sandtorstr. 1, 39106 Magdeburg
    Tel: +49 (0)391-6110-450
    E-mail: benner@mpi-magdeburg.mpg.de
Mitarbeiter:
  • Martin Köhler
    Max Planck Institut for Dynamics Complex Technical Systems Magdeburg,
    Computational Methods in Systems and Control Theory,
    Sandtorstr. 1, 39106 Magdeburg
    Tel: +49 (0)391-6110-445
    E-mail: koehlerm@mpi-magdeburg.mpg.de
  • Dr. Jens Saak
    Max Planck Institut for Dynamics Complex Technical Systems Magdeburg,
    Computational Methods in Systems and Control Theory,
    Sandtorstr. 1, 39106 Magdeburg
    Tel: +49 (0)391-6110-216
    E-mail: saak@mpi-magdeburg.mpg.de
Laufzeit: Seit 12/2010

Projektbeschreibung:
Der Rechen-Cluster otto stellt die neue zentrale Rechenressource des MPI Magdeburg dar. Er stellt den Wissenschaftlern genügend Rechenleistung für grosse und insbesondere verteilte Simulationen zur Verfügung. Unsere Arbeitsgruppe nutzt ihn für Berechungen und Simulationen, welche unseren anderen Forschungsaktivitäten entstammen.

Hardware

Es handelt sich bei otto um einen heterogenen Cluster. Wir haben 2 verschiedene Arten von Rechenknoten im Einsatz.
Knoten Typ 1 Knoten Typ 2
Prozessor: 2x Intel Xeon Westmere-EP X5650 2x AMD Opteron Magny Cours 6168
Kerne je Prozessor: 6 12
Taktfrequenz: 2.67 GHz 1.9 GHz
Arbeitsspeicher: 48 GB DDR3 ECC 96 GB DDR3 ECC
Festplatte: keine 600GB SAS
Netwerk: 40 Gbit/s Infiniband QDR 40 Gbit/s Infiniband QDR
Theoretische Leistung je Knoten: 128 GFlop/s 182 GFlop/s
Anzahl der Knoten: 90+1 16
Linpack Leistung aller Knoten: 10.36 TFlop/s 2.3 TFlop/s
Weitere Hardwarekomponenten sind:
  • 1 Login Konten auf Intel Xeon E5620 Basis
  • 1 Service Knoten auf Intel Xeon E5620 Basis
  • 1 Appliance auf Intel Xeon L3406
  • 28 TB Panasas PanFS Netzwerkspeciher
  • 5 Voltaire QDR Infiniband Switches

Software

Den Nutzern wird verschiedene Software aus dem Bereich Mathematik und Software Entwicklung zur Verfügung gestellt. Dies sind unter anderem:
  • GNU Compiler Collection 4.5.1
  • Intel C/C++/Fortran Compiler 12
  • Matlab 2010b
  • Gromacs 4.5
  • Sun Grid Engine 6.2
  • diverse Bibliotheken der numerischen Mathematik (BLAS, Lapack, Deal.II, Fenics)
Die installierte Software kann mit Hilfe des Modules-Systems ausgewählt und konfiguriert werden. Weitere Software wird entsprechend den Anforderungen der Wissenschaftler installiert.

Die Nutzung des Clusters erfordert eine separate Anmeldung. Weitere Informationen sind im internen Wiki des Institutes zu finden.


Ausgewählte Veröffentlichungen:

HPC en simulación y control a gran escala (in Spanish);
Peter Benner, Pablo Ezzatti, Hermann Mena, Enrique S. Quintana-Ortí, and Alfredo Remón;
; November 2011.
Solving Matrix Equations on Multi-core and Many-core Architectures;
Peter Benner, Pablo Ezzatti, Hermann Mena, Enrique S. Quintana-Ortí, and Alfredo Remón;
; MPI Magdeburg Preprint MPIMD/11-07, October 2011.
Efficiency improving implementation techniques for large scale matrix equation solvers;
Köhler, Martin; Saak, Jens;
CSC Preprint 09-10;
; 2009.
Parallel Implementation of LQG Balanced Truncation for Large-Scale Systems;
Badía, José M., Benner, Peter, Mayo Rafael, Quintana-Ortí Enrique S.; Quintana-Ortí, Gregorio; Remón, Alfredo;
; 2007.
Stabilizing Large-Scale Generalized Systems on Parallel Computers Using Multithreading and Message-Passing;
Peter Benner, Maribel Castillo, Rafael Mayo, Enrique S. Quintana-Ortí, and Gregorio Quintana-Ortí;
Concurrency and Computation: Practice and Experience, Vol. 19, No. 4, pp. 531-542, 2007.;
DOI: 10.1002/cpe.1148
Efficient Algorithms for Generalized Algebraic Bernoulli Equations based on the Matrix Sign Function;
Sergio Barrachina, Peter Benner, and Enrique S. Quintana-Ortí;
Numerical Algorithms, Vol. 46, No. 4, pp. 351-368, 2007.;
Parallel Solution of Large-Scale and Sparse Generalized Algebraic Riccati Equations;
Badía, José M., Benner, Peter, Mayo Rafael, Quintana-Ortí Enrique S.;
in W. E. Nagel, W. V. Walter, and W. Lehner: Euro-Par 2006 Parallel Processing: 12th International Euro-Par Conference  :  Lecture Notes in Computer Science, Vol. 4128, pp. 710-719;
Springer-Verlag, Berlin, Heidelberg, New York; 2006. ISBN/ISSN: 978-3-540-37783-2
DOI: 10.1007/11823285_74.
Solution of Discrete-Time Riccati Equations via Structure-Preserving Doubling Algorithms on a Cluster of SMPs;
Barrachina, Sergio; Benner, Peter; Quintana-Ortí, Enrique S.;
submitted for publication; 2006.
Parallel Algorithms for Balanced Truncation Model Reduction of Sparse Systems;
Benner, Peter; Quintana-Ortí, Enrique S.; Mayo, Rafael; Badía, José Manuel;
in J. Dongarra, K. Madsen, J. Wasniewski: Applied Parallel Computing: 7th International Conference, PARA 2004, Lyngby, Denmark, June 20-23, 2004. Revised Selected Papers  :  Lecture Notes in Computer Science, Vol. 3732, pp. 267-275;
Springer-Verlag; 2006. ISBN/ISSN: 3-540-29067-4
Original available from SpringerLink, DOI: 10.1007/11558958_32.
Balanced Truncation Model Reduction of Large and Sparse Generalized Linear Systems;
M. Badía, José; Benner, Peter; Mayo, Rafael; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio; Remón, Alfredo;
Chemnitz Scientific Computing Preprints 06-04, TU Chemnitz; 2006. ISBN/ISSN: 1864-0087
Parallel Order Reduction via Balanced Truncation for Optimal Cooling of Steel Profiles;
Badia, Jose M.; Benner, Peter; Mayo, Rafael; Quintana-Ortí, Enrique; Quintana-Ortí, Gregorio; Saak, Jens;
in J.C. Cunha and P.D. Mdeiros (eds.): Euro-Par 2005 Parallel Processing: 11th International Euro-Par Conference, Lisbon, Portugal, August 30 - September 2  :  Lecture Notes in Computer Science, Vol. 3648, pp. 857-866;
Springer-Verlag; 2005.
see also SfB393 Preprint 05-02.
Parallel Solution of Large-Scale Algebraic Bernoulli Equations with the Matrix Sign Function Method;
Benner, Peter; Barrachina, Sergio; Quintana-Ortí, Enrique S.;
International Journal of Computational Science and Engineering (IJCSE); 2005.
Parallel Algorithms for Balanced Truncation of Large-Scale Unstable Systems;
Benner, Peter; Barrachina, Sergio; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio;
Proceedings of 44th IEEE Conference on Decision and European Control Conference  :  pp. 2248-2253;
ECC; 2005. ISBN/ISSN: 0-7803-9568-9
Solving Stable Sylvester Equations via Rational Iterative Schemes;
Benner, Peter; Quintana-Ortí, Enrique S., Quintana-Ortí, Gregorio;
Journal of Scientific Computing  :  
Original available from SpringerLink.; 2005.
Solving Linear-Quadratic Optimal Control Problems on Parallel Computers;
Benner, Peter; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio;
Preprintreihe des Chemnitzer SFB 393  :  Preprint 19 / 2005;
; 2005.
Parallel Model Reduction of Large-Scale Unstable Systems;
Benner, Peter; Castillo, Maribel; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio;
in G.R. Joubert, W.E. Nagel, F.J. Peters, W.V Walter: Parallel computing: Software Technology, Algorithms, Architectures & Applications, Advances in Parallel Computing, Vol. 13, pp. 251-258;
Elsevier B.V (North-Holland); 2004. ISBN/ISSN: 0-444-51689-1
Parallel Model Reduction of Large-Scale Linear Descriptor Systems via Balanced Truncation;
Benner, Peter; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio;
in M. Daydé, J.J. Dongarra, V. Hernández, J.M.L.M. Palma: High Performance Computing for Computational Science - VECPAR 2004  :  Lecture Notes in Computer Science, No. 3402, pp. 340-353;
Springer-Verlag, Berlin, Heidelberg; 2004. ISBN/ISSN: 3-540-25424-2
State-Space Truncation Methods for Parallel Model Reduction of Large-Scale Systems;
Benner, Peter; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio;
Parallel Computing  :  Vol. 29 (special issue on "Parallel and Distributed Scientific and Engineering Computing"), pp. 1701-1722;
; 2003.
Parallel Algorithms for Model Reduction of Discrete-Time Systems;
Benner, Peter; Quintana-Ortí, Enrique S.; Quintana-Ortí, Gregorio;
International Journal of System Sciences  :  Vol. 34, No. 5, pp. 319-333;
; 2003.
The retrievable version appeared as Berichte aus der Technomathematik, Report 00-18, December 2000..
Numerical Solution of Discrete Stable Linear Matrix Equations on Multicomputers;
Peter Benner, Enrique S. Quintana-Ortí, and Gregorio Quintana-Ortí;
Parallel Algorithms and Applications (since 2005: International Journal of Parallel, Emergent and Distributed Systems) Vol. 17, No. 2, pp. 127-146, 2002.;
DOI: 10.1080/10637190208941436
Free preprint available at Berichte aus der Technomathematik, Report 99-13, November 1999.
Parallel Algorithms for LQ Optimal Control of Discrete-Time Periodic Linear Systems;
Peter Benner, Ralph Byers, Rafael Mayo, Enrique S. Quintana-Ortí, and Vicente Hernández;
Journal of Parallel and Distributed Computing, vol. 62, pp. 306-325, 2002.;
The retrievable version appeared as Berichte aus der Technomathematik, Report 01-02, February 2001.
Efficient Numerical Algorithms for Balanced Stochastic Truncation;
Peter Benner, Enrique S. Quintana-Ortí, and Gregorio Quintana-Ortí;
International Journal of Applied Mathematics and Computer Science, Vol. 11, No. 5, pp. 1123-1150, 2001.;
The retrievable version appeared as Berichte aus der Technomathematik, Report 01-03, March 2001.
Balanced Truncation Model Reduction of Large-Scale Dense Systems on Parallel Computers;
Peter Benner, Enrique S. Quintana-Ortí, and Gregorio Quintana-Ortí;
Mathematical and Computer Modelling of Dynamical Systems, Vol. 6, No.4, pp. 383-405, 2000.;
The retrievable version appeared as Berichte aus der Technomathematik, Report 99-07, September 1999.

 Top
Jens Saak, Martin Köhler, saak, koehlerm@mpi-magdeburg.mpg.de