The SCaLeS Report Opportunities and Needs in Basic Energy Sciences

The SCaLeS ReportThe SCaLeS ReportOpportunities and Needs in Basic Energy Opportunities and Needs in Basic Energy SciencesSciences

Thom H. Dunning, Jr.

Joint Institute for Computational SciencesUniversity of Tennessee • Oak Ridge National Laboratory

Oak Ridge, Tennessee

Outline of PresentationOutline of Presentation

BackgroundTrends: Computing Technologies

Trends: Scientific Applications

Scientific Opportunities

SCaLeS Workshop

SCaLeS ReportReports, Editors, and Process

Recommendations

BackgroundBackground Trends: Computing TechnologiesTrends: Computing Technologies

ComputersMicroprocessor performance continuing to double every 18-24 months, but …

increasing mismatch with memory subsystem performanceincreasing mismatch with communications subsystem performance

StorageDisk storage capacity doubling every year, but …

data transfer rates increasing only modestly

Communications FabricIncreasing performance, but …

increasing mismatch with performance of computational nodesincreasing mismatch with needed I/O transfer rates

BackgroundBackground Trends: Scientific ApplicationsTrends: Scientific Applications

Computational ModelsContinually refining existing models and creating new models

Multi-physics and multi-scale problems pose challenges

Parallel ComputingIncreasing use of parallelism

Most codes scale to 10s of processors, a few to 1-2,000 processors, but almost none to 10,000 processors

Mathematical TechniquesNew approaches hold great promise

Linear scaling reducing the growth rate in computational cost with increasing molecule size

Scientific OpportunitiesScientific Opportunities

Combustion ScienceReacting chemical flows

Autoignition

Molecular ScienceChemical reactivity (combustion, catalysis)

Heavy-element chemistry

Materials ScienceMaterials design

Multiscale materials modeling

NanoscienceSelf-assembly

Simulation of nano-devices

SCaLeS WorkshopSCaLeS Workshop

Date: June 23-24, 2003

Location: Arlington, Virginia

Organizer: D. Keyes, Columbia University

Goal: to assess the major opportunities and challenges facing computational science in areas of strategic importance to the Office of Science

Participants: 300+ scientists and engineers from academia, national laboratories, federal agencies and other institutions

SCaLeS ReportSCaLeS Report

EditorsDavid Keyes, Editor-in-Chief

Phil Colella, LBNL (mathematics); Thom Dunning, UT/ORNL (science); Bill Gropp, ANL (computer science)

Topical EditorsChemistry: R. Harrison, ORNL; T. Windus, PNNL

Combustion: J. Bell, LBNL; L. Rahn, SNL

Materials Science: F. Gygi, LLNL; M. Stocks, ORNL

Nanoscience: P. Cummings, Vanderbilt; L-W. Wang, LBNL

ProcessPreliminary topical reports compiled from Workshop notes

Reports iterated with Workshop participants plus others

SCaLeS Report SCaLeS Report ((cont’dcont’d))

Two VolumesVolume 1. Summary and recommendations

Available for download: http://www.pnl.gov/scales/

Volume 2. Detailed discussion of scientific opportunities and challenges

Available early next year

SciDAC: Successful Prototype to Build SciDAC: Successful Prototype to Build OnOn

Software Infrastructure

SCIENTIFIC

CODES

SI

MULATION

OPERATING

SYSTEM

DATAGRIDS

COLLABORATORIES

MATHEMATICS

Data Analysis &Visualization

Scientific DataManagement

Problem-solvingEnvironments

ProgrammingEnvironments

COMPUTING SYSTEMSSOFTWARE

HardwareInfrastructure

QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture.

QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture.

QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture.

SCaLeS ReportSCaLeS Report

RecommendationsRecommendations

Investments in Foundations of Computational Modeling and Simulation

#1. Computational Science

#5. Basic Theory and Mathematical Algorithms

#6. Recruit Computational Scientists

Investments in Hardware and Software Infrastructure#2. Multidisciplinary Teams

#4. Computing Systems and Scientific Applications Software

#3. Capability and Capacity Computing

#8. New Computer Architectures for Scientific Computing

Investments in Networking and Collaboration Technologies

#7. Network Infrastructure and Software to Support Distributed Computing and Data Resources and Scientific Teams

SCaLeS ReportSCaLeS Report

Recommendations: Foundations Recommendations: Foundations

Recommendation #1

Major new investments in computational science are needed in all of the mission areas of DOE’s Office of Science, so that the United States is the first, or among the first, to capture the new opportunities presented by the continuing advances in computing power.

Recommendation #5

Additional investments in hardware facilities and software infrastructure should be accompanied by sustained collateral investments in algorithm research and theoretical development.

Recommendation #6

Computational scientists of all types should be proactively recruited with improved reward structures and opportunities as early as possible in the educational process so that the number of trained computational science professionals is sufficient to meet present and future demands.

Investments in Computational ScienceInvestments in Computational Science

Advances in Molecular SimulationsAdvances in Molecular Simulations

Bond energies critical for describing many chemical phenomena

Accuracy of calculated bond energies increased dramatically from 1970-2000

Due to advances inTheoretical methodology

Computational techniques

Computing technology

1

10

100

1970 1980 1990 2000

Err

or

(kcal/

mol)

To achieve 1 kcal/mol accuracy:

CCSD(T) in 1989cc-Basis Sets in 1989Faster processors in 1990s

SCaLeS ReportSCaLeS Report

Recommendations: InfrastructureRecommendations: Infrastructure

Recommendation #2Multidisciplinary teams, with carefully selected leadership, should be assembled to provide the broad range of expertise needed to address the intellectual challenges associated with translating advances in science, mathematics and computer science into simulations that can take full advantage of advanced computers.

Recommendation #4Investment in hardware facilities should be accompanied by sustained collateral investment in the software infrastructure for them. The efficient use of expensive computational facilities and the data they produce depends directly upon multiple layers of systems software and scientific software which, together with the hardware, are the engines of scientific discovery …

Developing New Simulation Developing New Simulation CapabilitiesCapabilities

ComputerScience

AppliedMathematics

Theory(mathematical model)

Computational Science(scientific codes)

ComputationalPredictions

Experiment? YES

Basic MathAlgorithms

Computer SystemsSoftware

Prob

lem

with

Com

puta

tiona

l Met

hod?

Problem withMathematical Model?

NO

Inadequate

Adequate

Performance?

New Tool forScientific Discovery

SCaLeS ReportSCaLeS Report

Recommendations: InfrastructureRecommendations: Infrastructure

Recommendation #3Extensive investments in new computational facilities is strongly recommended, … New facilities should strike a balance between capability computing for those “heroic simulations” that cannot be performed in any other way, and capacity computing for “production” simulations that contribute to the steady stream of progress.

Recommendation #8Federal investments in innovative, high-risk computer architectures that are well suited to scientific and engineering simulations is both appropriate and needed to complement commercial research and development. The commercial computing marketplace is no longer effectively driven by the needs of computational science.

Branscomb ReportBranscomb Report

From Desktop to TeraflopFrom Desktop to Teraflop

FrontierComputers

Supercomputers

Mid-range ParallelComputers and Clusters

Personal Computersand Workstations

High-endCapacity Computing

WorkgroupCapacity Computing

Personal Computing

CapabilityComputing

Incr

easi

ng C

ost

per

Flo

p

Incr

easi

ng C

apab

ilit

y

Parallel Simulations: Hard Parallel Simulations: Hard vsvs Soft Soft ScalingScaling

Spee

d-up

Number of Processors

“Hard” Scaling

– near linear speed-up independent of problem size

– uncommon

“Soft” Scaling

– decreasing speed-up with constant problem size

– increase problem size to maintain scaling

but cost of calculation can increase more rapidly than that gained from increased scalability

– common

increasingproblem

size

SCaLeS ReportSCaLeS Report

Recommendations: Networks and Recommendations: Networks and CollabsCollabs

Recommendation #7

Sustained investments must be made in network infrastructure for access and resource sharing, as well as in the software needed to support collaboration among distributed teams of scientists, recognizing that the best possible science teams will be widely separated geographically and that researchers will generally not be co-located with facilities and data.

Distributed Teams and ResourcesDistributed Teams and Resources

WorkingTeam

High-speed networks plus grid and collaboratory software are needed to connect researchers with each other and with computing and data resources.

“The rising tide of change shows no respect for the established order. Those who are unwilling or unable to adapt in response to this profound movement not only lose access to the opportunities that the information technology revolution is creating, they risk being rendered obsolete by smarter, more agile, or more daring competitors.”

Jack J. DongarraUniversity of Tennessee

Advances in Computer Technology