Project

Back to overview

Computing equipment

English title Computing equipment
Applicant Rosenthal Joachim
Number 144973
Funding scheme R'EQUIP
Research institution Institut für Mathematik Universität Zürich
Institution of higher education University of Zurich - ZH
Main discipline Mathematics
Start/End 01.05.2013 - 30.04.2014
Approved amount 120'220.00
Show all

Keywords (1)

computing equipment

Lay Summary (German)

Lead
Computer werden mittlerweile in einigen Gebieten der Mathematik eingesetzt, um Vermutungen zu untersuchen, sowie Informationen über das praktische Verhalten von Algorithmen und Näherungslösungen von Systemen zu sammeln.Unser bisheriger Berechnungsserver ist seit längerer Zeit völlig ausgelastet. Einige geplante Berechnungen und Simulationen konnten deswegen gar nicht oder stark verzögert durchgeführt werden. Die neuen Server werden den Bedarf für die kommenden drei Jahre decken.
Lay summary
Inhalt und Ziel des Forschungsprojekts

Drei Forschungsprojekte, die ohne die zusätzliche Rechenleistung der neuen Server nur sehr eingeschränkt möglich wären, sind:
1) Die experimentelle Untersuchung von Gitterreduktion;
2) Statistische Modellierung von grossen multivariaten nicht-stationären spatio-temporalen Klimafeldern;
3) Die AL-Basis zur Lösung von elliptischen Problemen in heterogenen Körpern.


Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojekts

Giterreduktionsalgorithmen werden unter anderem in der Kodierungstheorie und Kryptographie verwendet.

Die Untersuchung der inhärente Unsicherheit von globalen Klimamodellen hat durch den Einfluss des Klimawandels auf Ökosystem, Wirtschaft und möglicherweise das Gesundheitswesen wichtige Auswirkungen.

Elliptische Differentialgleichungen haben Anwendungen bei der Untersuchung von zusammengesetzen und porösen Materialien, und der Untersuchung turbulenten Transportes.
Direct link to Lay Summary Last update: 13.03.2013

Responsible applicant and co-applicants

Collaboration

Group / person Country
Types of collaboration
Reto Knutti, Institute for Atmospheric and Climate Science, ETHZ Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Michael Schneider, TU Darmstadt Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Claus-Peter Schnorr, University of Frankfurt Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Stephan R. Sain, Institute for Mathematics Applied to Geosciences, NCAR, Boulder United States of America (North America)
- in-depth/constructive exchanges on approaches, methods or results

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Seminar, Department of Statistics, Kansas State University Individual talk An asymptotic framework for multivariate tapering 25.04.2014 Manhattan, KS, United States of America Furrer Reinhard;
Spatial Statistics for Environmental and Energy Challenges Talk given at a conference Extending tapering to multivariate spatial processes: concepts and illustrations 12.03.2014 Thuwal, Saudi Arabia Furrer Reinhard;
Workshop On Coding and Information Theory (WCI 2013) Talk given at a conference Subspace Codes and Orbit Codes 11.12.2013 Hong Kong, Hongkong Rosenthal Joachim;
International Symposium on Numerics and Scientific Computing Talk given at a conference Composite Finite Elements 22.10.2013 Frankfurt, Germany Sauter Stefan;
Conference on Random network codes and Designs over GF(q) Talk given at a conference List Decoding of Subspace Codes 18.09.2013 Ghent, Belgium Rosenthal Joachim;
IAMG 2013 Talk given at a conference Statistical modeling of a former Arctic Ocean ice shelf complex using Antarctic analogies 02.09.2013 Madrid, Spain Furrer Reinhard;
SIAM Conference on Applied Algebraic Geometry Talk given at a conference List Decoding of Subspace Codes 01.08.2013 Colorado , United States of America Rosenthal Joachim;
Karlsruher PDE Seminar Individual talk Fast Numerical Solution of Hyperbolic Integral Equtions in Exterior Domains 04.07.2013 Karlsruhe, Germany Sauter Stefan;
RMMM 2013 Talk given at a conference The saturation property for p-refinement 01.07.2013 Jyväskylä, Finland Sauter Stefan;
MAFELAP 2013 Talk given at a conference Finite Elemts for Elliptic Eigenvalue Problems in the Preasymptotic Regime 11.06.2013 London, Great Britain and Northern Ireland Sauter Stefan;
MAFELAP 2013 Talk given at a conference A Generalized Convolution Quadrature with Variable Time Stepping 11.06.2013 London, Great Britain and Northern Ireland Sauter Stefan;
Workshop: Error Estimates and Adaptive Mesh Renement Strategies for Boundary Element Methods Talk given at a conference Quadrature Techniques Convergence and quasi-optimality for Boundary Element Methods 22.05.2013 Paris, France Sauter Stefan;
The Art of Iterating Rational Functions over Finite Fields Talk given at a conference Some Dynamical Systems over Finite Fields appearing in Coding Theory and Cryptography 05.05.2013 Banff, Canada Rosenthal Joachim;


Self-organised

Title Date Place
Swiss Numerics Day 25.04.2014 Zürich, Switzerland
Pro*Doc Summer Retreat 14.08.2013 Disentis, Switzerland

Knowledge transfer events

Active participation

Title Type of contribution Date Place Persons involved


Associated projects

Number Title Start Funding scheme
178738 Generalized Convolution Quadrature and Directional H2 Matrices for Solving a Wave Transmission Problem with an Application to Determine the Bedrock of Glaciers 01.11.2018 Project funding
129782 Multivariate non-stationary spatial mixed-effects models for intelligent continuous compaction control 01.04.2010 Project funding
137146 Efficient Numerical Methods for Partial Differential Equations 01.10.2011 ProDoc
143282 Multivariate modeling of large non-stationary spatial and spatio-temporal climate fields 01.01.2013 Project funding
138080 Algebraic Constructions and Decoding of Network Codes 01.10.2011 Project funding

Abstract

In several areas of mathematics, both in pure and applied mathematics, computer experiments play an important role in research. Computers allow to compute larger and more complex examples than are feasible with just pen and paper, allow to verify or disprove conjectures, gain information on the practical behavior of algorithms, approximate solutions to various kinds of (systems of) equations, etc.The institute's current compute server is a Fujitsu RX 900 with 64 cores running at 2.0 GHz with 512 GB RAM. While this machine provides a lot of computing power, it is used to capacity most of the time. Most of the usage is generated by members of the working groups of Professors J. Rosenthal, R. Furrer and S. Sauter. In fact, the compute server does not provide enough computing power for all planned computations, simulations and experiments.Another problem is that the centralization of all computations on one server is sometimes problematic. First, hardware defects such as a CPU or memory defect result in the whole machine being shut down, which means that all running jobs are killed. Moreover, in case the running programs manage to exhaust the main memory, the resulting swapping process essentially brings the server to a grinding halt, which has to be resolved by a reboot.Another effect is that different processes affect each others, for example by thrashing the cache. Several processes, or one process with several threads, which manages to trash the caches of several CPUs can slow down other processes quite extremely. Since such effects are not counted when measuring CPU time, this can ruin timing experiments of algorithms.To improve on this situation, and also to increase stability, we propose to extend the computing machinery to - four new twin servers for fast computations with little interference, resulting in eight independent servers with 12 cores and 192 GB RAM each; - one large shared memory machine with 80 cores and 2 TB RAM; and - a local storage solution providing fast access to 24 TB of gross mirrored disk space which will be made usable with RAID-Z.The eight smaller machines are supposed to host jobs which require a large amount of memory and run for a long time, or jobs which should better be run on an own machine to avoid unwanted interference from other processes. The one large machine is supposed to be used for many processes, or for experiments with a very large number of cores per process or huge main memory requirements.This new hardware will allow a wide range of new computations, simulations and experiments to be run. This is true in particular for the projects listed in the research plan, which will be the main uses for the new machines. These projects cover important topics such as lattice reduction and statistical modeling of climate fields. Besides these projects, the hardware will be of potential use for other projects, both at the Institute of Mathematics at the University of Zurich as well as for other institutes or working groups from other universities in Switzerland (compare the research plan). Until now, all our compute servers have also been used by members from different institutes and universities in Switzerland.
-