Participants have two sessions to present their latest
on-the-edge work to a panel of judges who represent
industry, universities and government laboratories. Groups
will be asked to give a poster presentation at the Gala
Opening on Monday night and a 20-minute demonstration to the
judging panel on Wednesday. Awards will be presented to the
winners on Thursday afternoon at the Awards Session.
DUAL-LEVEL PARALLEL ANALYSIS OF HARBOR
WAVE RESPONSE USING MPI AND OPENMP
Primary Contact: Henry A. Gabb,
gabb@ibm.wes.hpc.mil
Steve W. Bova, Mississippi State University; Clay P.
Breshears, Rice University; Henry A. Gabb and Christine
Cuicchi, Waterways Experiment Station DoD Major Shared
Resource Center; Richard Strelitz, Science Applications
International Corporation; Zeki Demirbilek, Waterways
Experiment Station Coastal and Hydraulics Laboratory
The application, CGWAVE, models harbor response taking
into account outside sea state, harbor shape and man-made
structures (i.e., piers, breakwaters, naval vessels). It is
a forecasting and nowcasting tool used in coastal and
military planning and civil engineering. Historically, a
lack of computing power has forced approximations that
limited the predictive capability of the model. We will show
how simultaneous MPI/OpenMP parallelism drastically improves
program performance and improves simulation accuracy.
METACOMPUTING THE EINSTEIN THEORY OF SPACETIME:
COLLIDING BLACK HOLES AND NEUTRON STARS ACROSS THE ATLANTIC
OCEAN
Primary Contact: Ed Seidel,
eseidel@aei-potsdam.mpg.de
Werner Benger,
Max-Planck-Institut-fuer-Gravitationsphysik,
Konrad-Zuse-Institut; Bernd Bruegmann,
Max-Planck-Institut-fuer Gravitationsphysik; Ian Foster and
Olle Larsson, Argonne National Laboratory; Joan Masso,
University of the Balearic Islands; Mark Miller, Washington
University; Jason Novotny, NCSA; Marcus Pattloch,
DFN-Verein; Edward Seidel and John Shalf, NCSA; Warren
Smith, Argonne National Lab; Wai-Mo Suen and Malcolm Tobias,
Washington University
Using tightly coupled supercomputers in Europe and
America, we propose to perform an intercontinental,
distributed simulation of the full 3D Einstein equations of
general relativity, calculating the collision of black holes
and neutron stars. The simulation itself will be distributed
across machines on both continents, utilizing Globus, and
will be controlled and displayed live on an Immersadesk
Virtual Reality system at SC98. The domain decomposition
will involve one compact object (either a neutron star of
black hole) in Europe and one in America. Fully utilizing a
transatlantic ATM network, the two objects will collide and
merge (in a virtual space "somewhere" over the Atlantic
Ocean).
A DYNAMIC MASTER-SLAVE APPROACH FOR THE SIMULATION OF
DISPERSED MULTIPHASE FLOWS
Primary Contact: Bernard Bunner,
bunner@engin.umich.edu
Bernard Bunner and Gretar Tryggvason, Department of
Mechanical Engineering and Applied Mechanics, University of
Michigan, Ann Arbor
Simulations of dispersed multiphase flows present some of
the most difficult challenges in computational fluid
dynamics because of the existence of a deformable boundary
within the flow domain. A finite-difference/front-tracking
method has been developed to accurately calculate the motion
of bubbles and drops in a suspending fluid. The interface
between the two fluids is tracked explicitly by a moving
two-dimensional mesh superimposed on the fixed
three-dimensional computational grid on which the
Navier-Stokes equations are solved. A simple and efficient
master-slave approach is implemented to parallelize the
linked list structures describing the moving mesh while
domain decomposition is the natural choice to parallelize
the fixed grid data. This state-of-the-art method has
already produced revolutionary results yielding new insight
into the dynamics of multiphase flows.
INDUSTRIAL MOLD FILLING SIMULATION USING AN
INTERNATIONALLY DISTRIBUTED SOFTWARE COMPONENT
ARCHITECTURE
Primary Contact: Randall Bramley,
bramley@cs.indiana.edu
Martin Audet, Jean-Francois Hetu and Florin Ilinca,
Industrial Materials Institute, NRC, Quebec; Peter Beckman
and William F. Humphrey, Los Alamos National Laboratory;
Randall Bramley, Fabian Breg, Prafulla Deuskar, Shridhar
Diwan, Donald F. McMullen, Dennis Gannon, Madhusudhan
Govindaraju, John N. Huffman, Juan Villacis and Eric
Wernert, Indiana University&endash;Bloomington; Ian Foster
and Steve Tuecke, Argonne National Laboratory; Benoit Ozell,
Centre de Recherche en Calcul Applique, Montreal
This project will connect high performance software and
hardware systems to provide an integrated solution
environment for a 3D parallel finite element code modeling
industrial material processes such as casting and injection
molding. It will encompass problems of resource allocation
and management, component systems for problem-solving
environments, parallel solution of implicit finite element
systems on complex geometries, solution of large-scale
distributed systems of equations and 3D immersive
visualization. One goal of this project is to demonstrate a
complete, end-to-end high performance parallel solution of
problems of major importance to industry, all occuring in a
geographically distributed hardware and software
environment.
INNOVATIVE WIDE AREA APPLICATIONS ON THE GUSTO GRID
TESTBED
Primary Contact: Ian Foster,
foster@mcs.anl.gov
Joe Bester, Ian Foster, Joe Insley, Stuart Martin, Warren
Smith, Brian Toonen, Steve Tuecke and Gregor von Laszewski,
Mathematics and Computer Science Division, Argonne National
Laboratory; Karl Czajkowski, Steve Fitzgerald, Carl
Kesselman, Mei Su and Marcus Thiebaux, USC Information
Sciences Institute; Sharon Brunett, California Institute of
Technology; Russ Miller, SUNY Stony Brook; Steve Wang, Ian
McNulty and Mark Rivers, Advanced Photon Source, Argonne
National Laboratory
We showcase the Globus metacomputing toolkit and the
associated GUSTO testbed, the first large-scale realization
of a high-performance distributed computing infrastructure.
We do this by using Globus and GUSTO to perform three unique
computations, none of which would have been possible without
Globus services and GUSTO resources: (1) collaborative
online analysis of data from a microtomographic beamline at
the Advanced Photon Source at Argonne; (2) record-setting
distributed interactive simulation, using multiple
supercomputers; and (3) high-throughput computing for
crystallographic phase problems.
THE TERABYTE CHALLENGE: AN OPEN TESTBED FOR MANAGING,
MINING AND MODELING MASSIVE AND DISTRIBUTED DATA
Primary Contact: Robert Grossman,
grossman@uic.edu
Robert Grossman, Stuart Bailey, Simon Kasif, Don Mon,
Vijay Natarajan, Harinath Sivakumar, Gutti Srinath,
Guruvayurappan Subramanyam, Andriy Turinskiy, Mohamed Zaheer
and Suma Batchu, University of Illinois at Chicago; Robert
Hollebeek, Peter Buneman, Don Benton and Pavlos Protopapas,
University of Pennsylvania; David Rocke, University of
California at Davis; Ken Sevik, University of Toronto; Yike
Guo, Imperial College, London; Drew Baden, University of
Maryland at College Park; Peter Milne, Cooperative Research
Center for Advanced Computational Systems, Australia; Bernie
O'Lear, National Center for Atmospheric Research (NCAR);
Robert Grossman and Michael Cornelison, Magnify, Inc.; Bob
Hollebeek, HUBS, Inc.
The Terabyte Challenge Testbed (TCTB) is an open,
distributed testbed for experimental studies and
demonstrations involving data mining, predictive modeling
and data intensive computing. We will demonstrate software
tools to work with clusters of workstations, metaclusters
consisting of geographically distributed clusters and
superclusters consisting of geographically distributed
clusters connected with high performance networks. The focus
will be on mining large, massive and distributed data sets.
We will demonstrate a variety of applications, including
mining scientific and engineering data, medical and health
care data and business data.
LEGION: SEAMLESS, SECURE WIDE-AREA METACOMPUTING
Primary Contact: Marty Humphrey,
humphrey@cs.virginia.edu
Marty Humphrey, Andrew Grimshaw and the Legion team at
the University of Virginia
Legion is an object-based, metasystems software project
that provides the appearance of a single virtual machine
from a collection of geographically-dispersed, heterogeneous
resources. Applications will be shown that illustrate the
programming environment and run-time environment of Legion.
These applications, including hydrodynamic turbulence
simulation with remote visualization (MPI and DICE), gene
sequence matching and a parameter space study, will be
executed using hundreds of heterogeneous machines spread
across DoD, NPACI, UVa and our booth. A visual display will
be used to monitor application performance on the nationwide
Legion net. The emphasis in the accompanying discussion will
be on ease of use, performance and security.
EVERYWARE: COMBINING DISPARATE SOFTWARE
INFRASTRUCTURES FOR PERFORMANCE
Primary Contact: Rich Wolski,
rich@cs.ucsd.edu
Rich Wolski, UC San Diego, NPACI, University of Tennessee
at Knoxville; John Brevik, UC Berkeley; Alan Su, UC San
Diego; Neil Spring, UC San Diego, University of Washington;
Chandra Krintz, UC San Diego, University of Tennessee at
Knoxville; Graziano Obertelli, UC San Diego
Our goal is to demonstrate the ways in which various
distributed and metacomputing infrastructures such as
Legion, Globus, Condor and Java can be profitably combined
with "vanilla" Unix systems to solve large-scale
computational problems. Everyware will allow processes and
computational agents running on supercomputers,
workstations, personal computers and within web browsers to
be used in concert by a single application.
NEAR REAL-TIME IMAGING OF HUMAN BRAIN ACTIVITY
Primary Contact: Greg Hood,
ghood@psc.edu
Greg Hood, Chad Vizino and Joel Welling, Pittsburgh
Supercomputing Center; Doug Noll and Andy Stenger;
University of Pittsburgh Medical Center; Jana Asher,
Carnegie Mellon University
We will demonstrate near real-time imaging of brain
activity with a live human subject in an MRI scanner at the
University of Pittsburgh Medical Center. Data will be sent
as it is acquired to the Cray T3E at the Pittsburgh
Supercomputing Center for processing, and the results then
sent to the show floor for visualization in the PSC research
booth on an SGI Onyx with Infinite Reality Engine. Our
presentation will describe the nature of functional imaging
(fMRI), the methods that we use to reconstruct
three-dimensional volumes from the raw scanner output and
the prospective uses for this technology.
SEMI-TRANSPARENT SUPERCOMPUTING: DYNAMIC VOLUME
RENDERING ON REMOTE HPC SYSTEMS
Primary Contact: Gregory Johnson,
johnson@sdsc.edu
Gregory Johnson, Jon Genetti and Mike Gannis, SDSC, NPACI
Volume rendering of medical data produces accurate,
highly detailed images of internal anatomy not available by
other means. However, multi-gigabyte data such as that from
the Visible Human Project (NLM), exceeds the ability of
workstation class systems with respect to generating images
at the rates required for an effective exploration of the
subject.
SDSC researchers have developed a distributed direct
volume rendering system called the Massively Parallel
Interactive Rendering Environment (MPIRE). MPIRE includes a
Java applet through which the user configures the desired
rendering parameters, target MPP system and compute
resources. A scalable parallel rendering engine is then
automatically started on the MPP over the specified number
of nodes, the data loaded and an image created and sent back
to the Java applet. From there, the image is automatically
updated by the engine as the user modifies the camera
position, lighting, coloration or any other rendering
parameter.
