PowerPoint Presentation - Computing, Information, and Networking <http://www-internal.ornl.gov/cind>

1
Computer Science and Mathematics Division
Jeff Nichols, Director May 20, 2003
http//www.csm.ornl.gov/
2
Reminder - Center for Computational Sciences
DOEs Advanced Computing Research Testbed

• Serves as focal point for scientific research
  community as it adapts to new computing
  technologies
• Provides organizational framework needed for
  multidisciplinary activities
• Addressing software challenges require long term
  collaborations among disciplinary computational
  scientists, computer scientists, and applied
  mathematicians
• Provides organizational framework needed for
  development of community codes
• Implementing many scientific codes requires wide
  range of disciplinary expertise
• Develops and deploys systems based on science
  needs

• Critical Issue
• Provide capability computing for Science mission

TFs
120
Astrophysics Nano and Molecular Science
100
80
60
Climate Fusion Combustion
40
20
0
2000
2002
2004
2006
3
Computer Science and Mathematics J. A. Nichols,
Director C. Hamby, Secretary (Acting)
ADVISORY COMMITTEE Thom Dunning Distinguished
Scientist Jerry Bernholc Visiting Distinguished
Scientist Jack Dongarra Distinguished
Professor David Keyes Visiting Distinguished
Scientist
Complex Systems J. Barhen L. C.
Holbrook Y. Y. Braiman R. J. Carter C.
D'Helon2 W. R. Garrett C. W. Glover W. P.
Grice W. C. Grimmell N. Imam2 S. Jeon2 D. L.
Jung J. Kelly1 Y-H. Kim5 V. Kireev2 S. M.
Lenhart H. K. Liu Y. Liu2 L. E. Parker R.
Perez N. Peterfreund V. Protopopescu N.S.V.
Rao D. B. Reister N. Smith1 D. R. Tufano
Network and Cluster Computing G. A. Geist
L. M. Wolfe P. K. Agarwal2 D. A. Bauer1 D. E.
Bernholdt M. L. Chen K. Chanchio V. Choppella2 S.
P. Dickson J. J. Dongarra4 T. H. Dunigan W. R.
Elwasif C. Engelmann2 J. L. Hespen1 G. H.
Kora1 J. A. Kohl B. M. Leuthke J. L. Mugler2 T.
J. Naughton2 P. E. Pfeiffer3 J. A. Rome N. F.
Samatova S. L. Scott J. Schwidder I.-P. G.
Watkins1 T. Wilde W. R. Wing
OPERATIONS COUNCIL Finance K. R. Spence Human
Resources N. Y. Wright Technical Information
and Communications Specialist/ Facility 6025 B.
A. Riley Recruiting/ Staffing J. K.
Johnson Computer Security T. K.
Jones4 ESH/Facility 6010 6012 R. J.
Toedte4 Facility 3546 D. E. Bernholdt4 Quality
Assurance R. W. Counts4
Statistics and Data Science C. K. Bayne L. C.
Holbrook R. W. Counts E. L. Frome B. M.
Horwedel B. L. Jackson G. Ostrouchov L.
Pouchard D. Schmoyer D. A. Wolf
Climate Dynamics J. B. Drake L. M.
Wolfe M. L. Branstetter2 D. J. Erickson X.
Guo3 M. W. Ham J. L. Hernandez2 G.
Mahinthakumar3 M. Sayeed3 P. H. Worley
Computational Mathematics E. F.
DAzevedo T. S. Darland V. Alexiades3 V. K.
Chakravarthy2 Y-S Chan2 G. I. Fann Z. Gan2 L. J.
Gray R. J. Harrison3 D. Keyes4 A. K. Khamayseh M.
R. Leuze S. PannalaT. Yanai2
Computational Materials Science T. C.
Schulthess T. S. Darland G. A. Aramayo S. C.
Barhen1 J. Bernholc4 G. P. Brown2 T. Dunning4 R.
G. Endres2 M. A. Fuentes-Cabrera2 J. P. Hague2 B.
C. Hathorn2 T. Kaplan P. R. Kent2 T. A.
Maier5 V. Meunier2 M. B. Nardelli3 D. M.
Nicholson D. W. Noid P. K. Nukala Y. Osetskiy L.
Petit2 B. Radhakrishnan G. B. Sarma W. A.
Shelton S. Shevlin2 S. Simunovic A. V.
Smirnov2 B. G. Sumpter J. C. Wells X-G. Zhang J.
Zhong2
Systems and Operations A. S.
Bland A. D. Harris R. A. Alexander M. W.
Arnold T. O. Barron R. D. Burris S. M. Carter F.
M. Fowler N. R. Hathaway T. K. Jones K. D.
Matney D. E. Maxwell D. L. Million G. G. Pike S.
Shpanskiy D. A. Steinert S. R. White V. L.
White (11) SAIC staff
Computational Biology G. A. Geist
(acting) L. M. Wolfe P. K. Agarwal2,4 P.
Chandramohan2 H.-S. Choi1 R. M. Day1 T.
Fridman A. A. Gorin M. R. Leuze4 R. V.
Munavalli1 G. Ostrouchov4 B.-H. Park2 L. C.
Pouchard4 N. F. Samatova4 Y. Xu G.-X. Yu2
Scientific Applications Support A. S.
Bland A. D. Harris M. R. Fahey R. A. Fahey C.
P. Halloy R. J. Toedte J. B. White, III K. L.
Wong
1 Student 2 Post doc 3 Joint Faculty 4 Secondary
Assignment 5 Wigner Fellow
05/06/03
4
Typical Application Development Team
Application Domain
e.g., HPF, MPI
e.g., BLAS, LAPACK
Computer Science
Applied Math
Core Technologies
5
Ideal Application Development Team
CORE
Application Domain
OSCAR, CCA, Global Arrays
PEIGS
CS
Math
Technologies
6
CSMD Provides Organizational Infrastructure for
Multidisciplinary Teams
Brings theoretical and computational application
scientists, computer scientists, and applied
mathematicians, together in long term
collaborations.
Applied Math
Computer Science
Theoretical Computational Domains
Goals Achieved
7
SciDAC Programmatic ElementsScientific
Discovery through Advanced Computing
BES, BER FES, HENP
ASCR
8
Systems and Operations

• Provides 24/7 operations coverage for the CCS
  high performance computing and storage systems
  and the ORNL mass storage systems.
• Provides systems administration, configuration,
  and management for the CCS high performance
  computing systems and for CSM workstations and
  servers.

• Hardware
• IBM Power4 (Cheetah)
• Cray X1 (Phoenix)
• SGI Altix
• IBM SP (Eagle)
• Compaq SC 64-node (Falcon)
• Compaq SC 16-node (Colt)
• Probe storage testbed
• eXtreme TORC (XTORC cluster)
• HPSS

9
Scientific Applications Support

• Provides consulting for users of the Center for
  Computational Sciences
• Administers the account approval process for the
  CCS and other CSM computing systems.

• Research Communities
• Astrophysics
• Biology
• Chemistry
• Climate
• Fusion
• Materials

10
Network and Cluster Computing
Significant impact and world-wide influence on
Parallel computing and the Science enabled by it

• Track record of developing
• very popular software
• PVM 400,000
• OSCAR 100,000
• Cumulvs - 300
• Influencing Standards
• MPI, BLAS, LAPACK, PAPI
• Enabling Science
• PVM, MPI, enote, etc.
• are widely used in
• education, research, and industry

Goal is to accelerate the process of Scientific
Discovery
11
Computational Mathematics

• Boundary Integral Method
• Functionally Graded Materials (FGM)
• Crack propagation
• Crystal growth
• Numerical linear algebra
• Problem specific preconditioners
• Parallel and out-of-core linear solvers for large
  dense matrices.
• Mesh generation
• Solution adaptive refinement
• Complex geometry
• Quasi-conformal mapping
• Multiphase CFD
• Spray combustion
• Large Eddy Simulation
• Fluidized bed
• Multiscale
• MRA, Multigrid, Multipole

12
Statistics and Data Science

• Applied Statistics
• Environmental Technology
• Environmental Monitoring
• Forensic Analysis
• Graphite Reactor Technology
• Tobacco Smoke Studies
• Tank Closure
• Future Combat Systems
• Chemical/Biological Warfare
• Biostatistics
• Epidemiology Surveillance
• Beryllium Lymphocyte Test

• Computational Statistics
• Distributed Data
• Data Mining Tools
• Spatial Statistical Models
• Database
• Atmospheric Radiation (ARM)
• Distributed Active Archive
• CO2 Database
• Mouse Tracking (MuTrack)
• Genome Database
• Earth Science Grid/Ontology

13
Computational Materials Science

• Materials science from traditional to
  nanoscience
• Continuum mechanics
• Microstructure / phase stability
• Nano-magnetism / magneto transport (lead theory
  of GMR)
• Carbon nanotubes / quantum dots
• High performance computing (fastest applications)
• Polymer chemistry
• Density functional theory
• Multi-scale modeling

• Diversified research
• Programmatic (BES ASCR/MICS EE)
• Industrial (CRADAs ranging from steel/aluminum
  industry to electronics / magnetic recording
  industry)
• Other DARPA ONR
• Relevance to other divisions MC / Solid State /
  Chemistry / Physics
• Center for Materials Research (CCS-CMR)

14
Electronic properties and Quantum Conductance in
intramolecular Carbon Nanotube Networks

• Carbon nanotubes are quasi 1D quantum wires that
  can sustain a ballistic electronic current,
  therefore, they are expected to be a constitutive
  component for the rapidly emerging field of
  molecular electronics.
• A branched junction is a necessary building block
  for nanotube network design. Three-way and
  four-way junctions have been observed
  experimentally.
• We show that the multi-terminal nanotube systems
  are ideal branching constituent for complex
  molecular-based nanoelectronics.

15
Climate Dynamics

• Numerical methods research for atmospheric
  dynamics
• Parallel algorithms for spectral methods
  (available in NETLIB)
• High resolution model development
• Benchmarking and performance characterization
  (instrumented software)

A Standard Test Set for Numerical Approximations
to the Shallow Water Equations in Spherical
Geometry, JCP, 1992
Evaluation of Early Systems http//www.csm.ornl.g
ov/evaluation
16
Simulation of Global DMS Flux from the Ocean
David Erickson(ORNL), Jose Hernandez(JICS), Matt
Maltrud(LANL), Shaoping Chu(LANL)

• The global ocean ecosystem model in POP developed
  at LANL has been coupled with the air-sea flux
  parameterization model at ORNL and the flux has
  been computed.
• The globally integrated flux of DMS from the
  ocean to the atmosphere is 23.8 Tg S yr-1, in
  reasonable agreement with experimentally
  developed estimates.
• The high resolution of the ocean DMS maps is
  evident in the spatial structure of DMS flux.
• This study required the computational capacity of
  the Center for Computational Sciences and was
  supported by the SciDAC CCSM Consortium Project.

A key climate interaction of an active
biochemical ocean and atmosphere is through the
sulfur cycle. The global flux of DiMethyl
Sulfide(DMS) from the ocean to the atmosphere is
shown as an annual mean Units are of mMol DMS
m-2 day-1.
17
Computational Biology

• Genome-scale modeling of biochemical pathways
• Extreme pathways (EP)defined within the context
  of convex analysis
• Parallel out-of-core EP algorithm
• Combinatorial Assignment Procedure (CAP)
• Combines robust mathematical algorithm of the
  assignment verification and combinatorial
  computational methods,
• fully automated resonance assignment.
• New hierarchical SVM-based classifier for
  high-throughput functional characterization
• Support Vector Machines (SVMs)
• Intelligent feature selection
• Hidden Markov Models (HMMs)
• 6 out 148 sequences were misclassified by
  cross-validation
• Able to discriminate functionally distant but
  very homologous genes
• More than 2000 functional groups are supported
• Petascale Distributed Data Analysis
• RACHET High Performance Framework for Distributed
  Data Analysis

18
Knowledge Fusion-based Mining for
Protein-Protein Interactions in GTL

• Characterization of Protein Machines is one of
  the key goals of the DOE Genomes To Life (GTL)
  Program
• ORNL has developed categorical data analysis for
  merging information from experimental and
  sequence databases (DIP, InterPro, Pfam)

• Algorithm achieves 90 prediction accuracy
• It provides strong candidates for detailed
  studies of protein-protein interface
• It provides building blocks for construction of
  protein-protein interaction networks

19
Complex SystemsFY 2002 Funding 8.6 9.1
M 0.9 M CapEq
MDA 2.0 M - sensitivity and uncertainty
analysis - remote sensing - POET - JTAMDO
INDUSTRY 0.4 M - Nicolet noninvasive
diagnostics - Caterpillar cooperative
robotics - TurboWave network protocols
NASA 0.4 M - deep-space
communications - strategic technologies
investment
DOE/SC/BES 2.5 M
CapEq 0.9 M - laser arrays -
complex patterns - nanoscale engineering -
quantum optics - nanobiomaterials -
information fusion

• Other DOD 0.5 M
• ONR sensors
• control of chaos
• - OSD directed microwave
• energy

Other DOE 0.9 M - ASCR computer
networks - FE hyperspectral remote
sensing
LDRD/SF 0.5 M - computer networks -
computational biology - quantum optics
Intelligence Community 0.4 M ( 0.5 M
pending) - NSA quantum cryptography - ARDA
photonic nanosystems - AGA classified - NRO
petascale optical computing
DARPA 1.0 M - robotics/urban warfare -
networks/ cyberwarfare - future combat systems
20
Control of Friction at the Nanoscale
Yehuda Braiman Jacob Barhen
Vladimir Protopopescu
Center for Engineering Science Advanced
ResearchComputer Science and Mathematics Division

• Friction is omnipresent in scientific,
  engineering, and technological applications
• New algorithm developed
• provides a revolutionary advance in control of
  friction in nanosystems
• fast and efficient
• enables to induce any arbitrarily chosen behavior
  compatible with the system's dynamics. 
• Methodology is based on two original
• concepts
• non-Lipschitzian dynamics
• global behavior targeting 
• Reference
• Physical Review Letters, 90 (in press 2003)

Figure Performance of control algorithm for four
values of the center of mass velocity ( 0, 0.5,
1.0, and 3.0) for a 15 - particle array. Control
was initiated at t2000. Blue lines show time
series of the center of mass velocities, while
red lines show the control. In all cases, the
desired behavior was rapidly achieved. All the
units are dimensionless and initial conditions
were chosen randomly.
21
Compact stellarator science discoveries have been
enabled by CCS computers
Physics/engineering based parallel optimizations
used to synthesize innovative compact QPS
stellarators
Massively parallel particle simulations for
physics analysis
Stellarator Research (ORNL, PPPL, Spain, U. of
Montana) - Recent Physical Review Letter
improved QPS reactor concept - Successful ORNL
facility design (QPS)
Magnetic field strength (Tesla)
22
Advanced Methods for Electronic Structure
R.J. Harrison, G.I. Fann, T. Yanai (ORNL) withG.
Beylkin (U. Colorado)

• Objectives
• Eliminate basis set error
• Correct scaling with system size
• Demonstrated for one-electron methods (DFT/HF)
• Investigation of two-electron models commencing
• Distinguishing features
• Multiresolution in multiwavelet bases
• Near-spectral precision
• Integral formulation of quantum theory
• efficient, avoids iterative solution of linear
  equations and problems with differential
  operators
• Separated representation of integral operators
• computation in higher dimensions
• Non-standard form of operators
• It is the first time that most of these methods
  have been employed in a 3-D application ... ...
  work in 6-D is now commencing.

23
(No Transcript)