U.S. Department of Energys Office of Science

1
U.S. Department
of Energys
Office of Science
Advanced Scientific Computing Research Planning
for the Future

• Basic Energy Sciences Advisory Committee

Dr. Daniel A. Hitchcock October 21,
2003 Daniel.Hitchcock_at_science.doe.gov
2
ASCR Program Overview
Research to enable
Applications BES, BER, FES, HEP, NP

• Nanoscience
• Materials
• Chemistry
• Combustion
• Accelerator
• High energy Physics
• Nuclear physics
• Fusion
• Climate
• Astrophysics
• Biology

• Network Environment
• Scientific Applications
• Genomes to Life

• Applied Mathematics
• Computer Science

• Grid enabling research

• Nanoscience

SciDAC!
Integrated Software Infrastructure Centers
(Mathematicians, computer scientists, application
scientists, and software engineers)

Next Generation Architecture
High Performance Computing and Network Facilities
for Science
3
Workshops and Reports

• Genomes to Life, 2001
• http//www.doegenomestolife.org/pubs.html
• Theory and Modeling in Nanoscience, May 2002
• http//www.sc.doe.gov/bes/besac/Theory20and20Mod
  eling20in20Nanoscience.pdf
• Blueprint for Future Science Middleware and Grid
  Research and Infrastructure, August 2002
• http//www.nsf-middleware.org/MAGIC/default.htm
• High Performance Network Planning Meeting, August
  2002
• http//doecollaboratory.pnl.gov/meetings/hpnpw/fin
  alreport/
• Fusion Simulation Project, December 2002
• http//ofes.fusion.doe.gov/News/FSP_report_Dec9.pd
  f
• DOE Science Network Meeting, June 2003
• http//gate.hep.anl.gov/may/ScienceNetworkingWorks
  hop/
• DOE Science Computing Conference, June 2003
• http//www.doe-sci-comp.info
• Science Case for Large Scale Simulation, June
  2003
• www.pnl.gov/scales/
• Workshop on the Road Map for the Revitalization
  of High End Computing
• http//www.cra.org/Activities/workshops/nitrd/
• Cyberinfrastructure Report

4
Strategic Issues for ASCREnabling World
Leadership in Science

• Providing high performance computing and network
  facilities
• Maintaining world-class research effort in
  applied mathematics, computer science, and
  computer networks
• Effective partnerships with applications
  scientists in all of the Offices in SC
• Effective partnerships with other Federal
  Agencies
• Accelerating transition from research to
  application and
• Long-Term support of software.

5
Research to Enable New Frontiers in Science
though Simulation

• The mathematics of complex and multiscale
  systems
• Ultrascalable algorithms for petascale systems
• The computer science to enable advanced
  computers and
• The computer science to transform petabytes of
  data into knowledge.

6
Compute Facilities and Testbeds
Definitions

• Experimental
• Proof of concept
• Small scale research projects
• Research and Evaluation (RE )Prototype
• Sufficient scale to enable evaluation of
  scientific potential
• Research Projects
• Enables new science for a few brave users
• High Performance Production Capability
• Stable, multi-user capability environment
• Large user support, consulting and training
  investment
• Direct support of agency mission
• Leadership-Class
• Most Capable System available for a class of
  applications
• Small number of projects
• Resource for national science community, managed
  similar to light source or high energy physics
  facility

7
Compute Facilities and TestbedsStrategy

• Maintain Current Investment in High Performance
  Production Capability for SC Missions
• Increase number of RE Prototypes
• Not all funded by ASCR
• Reduce architecture risk
• Manage as research projects
• Improve coupling with vendors in design stage
• Couple software research to facility and testbed
  operations
• Leadership-Class Compute Capability for Open
  Science and
• Tighten coupling between RE Prototype Evaluation
  and High Performance Production Capability and
  Leadership-Class

8
Compute Facilities and TestbedsTimeline
Experiment
RE Prototype
Leadership-Class
SOFTWARE
High Performance Production Capability
High Performance Capacity
Desktop
9
Network Environment Research
Science applications and specialized experimental
facilities are n-way interconnected to terascale
computing, petascale storage, high-end
visualization, and remote collaborators in a
seamless environment that provides the
performance level required to move science,
especially large-scale science, to a new
regimerapid scientific progress through the
interplay of theory simulation, and experiment.
10
Research to Enable New Frontiers in Science
though Networks

• End-to-end performance
• High-Performance Middleware
• Integrated testbeds and networks

11
Network Facilities and TestbedsStrategy

• Maintain Current Investment in High Performance
  Production Capability for SC Missions
• Develop RE Prototype Network
• Coordinated with other Agencies
• Rely on other Agencies for fundamental Hardware
  Research
• Focus on High Throughput Streams
• Network Measurement and Diagnosis
• Couple software and applications research to
  network facility and testbed operations and
  planning and
• Tighten coupling between RE Prototype Evaluation
  and High Performance Production Network.

12
Network Facilities and TestbedsTimeline
Experiment
Network Environment Applications
RE Prototype
High Performance Production Networks
13
Effective Partnerships with Applications
Scientists Partnerships Strategy

• Build on SciDAC
• Partnerships are critical to ASCR
• Increase joint Program Office collaboration to
  plan partnerships
• SciDAC
• Genomes to Life
• Nanoscience
• Fusion Simulation Project
• ASCR funds math, computer science and glue
  while Program Offices fund science

14
Effective Partnerships with Application
Scientists Long Term Support of Software Strategy

• Critical issue for delivery of software resulting
  from ASCR research as well as application
  community codes
• Must be coupled to research but is not research
• Issues
• Can be very expensive
• Low volume niche market
• Distributed benefit, centralized cost
• Many previous models have failed
• Hard to make a priority in a research
  environment.