HELP! * GREY=local Full HTML for

LOCAL foilset PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96

Given by Paul Smith DoE at JNAC (PetaFlops) Presentation on August 28,1996. Foils prepared August 25 1996
Abstract * Foil Index for this file

This was part of a set of PetaFlop (JNAC) Presentations to group of Federal Program Managers
JNAC = Joint National Advanced Computing Initiative
This summarizes results of the series of Petaflop Workshops and Conference Events
It Summarizes the case to move forward with the major JNAC Program

Table of Contents for full HTML of PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96

1 PetaFLOPS Workshops Results
2 PetaFLOPS Workshops Results
3 Purpose of Meeting "Why we're here?"
4 Steps to Progress
5 PetaFLOPS I
6 MAJOR FINDINGS PetaFLOPS I
7 MAJOR FINDINGS PetaFLOPS Bodega Bay Summer Study
8 MAJOR FINDINGS PAWS'96
9 MAJOR FINDINGS PetaSoft `96
10 SOME IMPORTANT RECOMMENDATIONS
11 PP Presentation
12 Conclusions -- need to polish this
13 The Message
14 Build on the Coherent R&D Planning from the PetaFLOPS Workshop Series
15 PP Presentation
16 The Message
17 JNAC
18 Near Term Recommentions
19 Next Steps

This table of Contents Abstract

HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 1 PetaFLOPS Workshops Results

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
DRAFT
The PetaFLOPS Workshops Series
August 28, 1996
Dr. Paul H. Smith
Department of Energy
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 2 PetaFLOPS Workshops Results

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
AGENDA
9:00 The PetaFLOPS Workshop Series Paul H. Smith
9:30 Drivers for High End Computing Paul Messina
9:45 Findings and Recommendations from Thomas Sterling
the PetaFLOPS Workshop Series
10:15 Break
10:45 Point Design Studies John Van Rosendale
11:00 Software Development Architecture Geoffrey Foz
11:15 Proposed JNAC Program Concept Rick Stevens
11:45 Near Term Steps and Discussion Paul H. Smith
12:30 Adjourn
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 3 Purpose of Meeting "Why we're here?"

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Brief major findings & recommendations of the PetaFLOPS workshops
  • the PetaFLOPS process
Identify the key drivers for advanced computational capabilities beyond HPCC
Recommend a program concept
  • Joint National Advanced Computing initiative
Discuss future actions necessary to mold & advocate a coherent successful R&D program
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 4 Steps to Progress

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * Critical Information in IMAGE
Full HTML Index
Approvals
Need
Feasibility
Program Plans
Implementation
Software
Workshop
June `96
Scope
Architecture
Workshop
April `96
Petaflops Frontier 2
October `96
Algorithms
Workshop
April `97
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 5 PetaFLOPS I

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
PetaFLOPS Challenge
  • The Workshop on Enabling Technologies for PetaFLOPS Computing
    • Held February 22 - 24, 1994 in Pasadena CA
      • 60 plus experts in all aspects of high-performance computing technology
      • To establish the basis for considering future research initiatives that lead to petaFLOPS scaled computing systems
    • New Technologies explored
    • Architecture
      • Computing & Memory dominant
      • Distributed Networked Systems
    • Set tone for subsequent Workshops
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 6 MAJOR FINDINGS PetaFLOPS I

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
PetaFLOPS is possible / feasible in 15 - 20 years
Applications will demand petaFLOPS (1015) performance
Memory is dominant implementation factor
Semiconductor estimates (SIA) allow for 1015 (FL)OPS
  • 20 years, up to $1 Billion, 60 Megawatts
  • Fits in room
Superconducting technology very promising for processor performance
Planned system software technology is inadequate to support petaFLOPS goals
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 7 MAJOR FINDINGS PetaFLOPS Bodega Bay Summer Study

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Accelerate PetaFLOPS goals to 10 years.
Multiple applications classes exist.
Technology transition (paradigm shift) required to get to PetaFLOPS.
Need to tie architectures closer to programming models
Role of algorithms must improve
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 8 MAJOR FINDINGS PAWS'96

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Accelerate PetaFLOPS goal to a decade.
Memory, memory and memory
  • latency
  • bandwidth
  • cost
New technologies (paradigm shifts)
  • superconducting chips
  • PIM chips
  • optical interconnects
Architectures
Algorithms
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 9 MAJOR FINDINGS PetaSoft `96

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Importance of software
Need for explicit, low-level control of hardware
Need for a layered software architecture
Existence of fundamental systems software scaling problems
Opportunities for immediate software effort
Interrelationship of hardware and low-level software
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 10 SOME IMPORTANT RECOMMENDATIONS

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Start S/W now (independent of H/W).
Develop layered architecture
Support strong R&D on new paradigm shift technologies
Conduct point design studies
  • in H/W and S/W.
Develop performance models
  • Multiple technology track demonstrations
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 11 PP Presentation

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * Critical Information in IMAGE
Full HTML Index
Steps Beyond TeraFLOPS
Year Peak Performance First Achieved
Floating Point Operations Per Second (FLOPS)
REYNOLDS-AVERAGED (N-A)
NAVIER-STOKES (N-S)
AIRCRAFT
Computational Needs
Rapid Path
Enhanced Evolution
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 12 Conclusions -- need to polish this

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
... well thought out process
... best experts in country
... joint research and development community
  • Academia
  • Private Sector
  • Federal Government
... vision of steps for the next 10 years
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 13 The Message

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Strong need for computing at the high end
PetaFLOPS levels of performance are feasible
Private sector is not going to do it
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 14 Build on the Coherent R&D Planning from the PetaFLOPS Workshop Series

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Develop Technical Approach --NOW
  • Elements, Phasing
  • Strategy for developing National Initiative
    • Community Support, New Start in FY 1998
Multi-agency efforts
DOE, NASA, NSA, NSF, DARPA
Pre-start in FY 1997
Consider current successful Federal and Industrial Models
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 15 PP Presentation

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
PetaFLOPS Workshops Results
DRAFT
Near Term Steps and Discussion
August 28, 1996
Dr. Paul H. Smith
Department of Energy
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 16 The Message

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Need to invest in computing at the high end
PetaFLOPS level of performance are feasible
Private sector is not going to do it
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 17 JNAC

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
Develop Technical Approach --NOW
  • Elements, Phasing
  • Strategy for developing National Initiative
    • Community Support, New Start in FY 1998
Multi-agency efforts
DOE, NASA, NSA, NSF, DARPA
Pre-start in FY 1997
Consider current successful Fed/Indust. Models:
  • DOE Battery consortia
  • SEMATEC
  • CRADA's
  • NISTs ATP.
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 18 Near Term Recommentions

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
....add major ones from the 4 workshops here
HELP! * GREY=local HTML version of LOCAL Foils prepared August 25 1996

Foil 19 Next Steps

From PetaFlop(JNAC) Overview Presentations -- Results of Studies and Next Steps Aug 28 96 JNAC (PetaFlops) Presentation -- August 28,1996. * See also color IMAGE
Full HTML Index
The PetaFLOPS Frontier (Oct 96)
  • Open community, international
PetaFLOPS Algorithms Workshop (Apr 97)
PetaFLOPS II Conference (Sept 97)
  • Algorithm driven Architecture
  • Buildability based on Point Designs against components
  • Applications drivers/scenarios
  • Leap Technology elements
Engage community in establishing challenges, directions, topics for research
Integrate into Federal R&D Strategic Planning
Northeast Parallel Architectures Center, Syracuse University, npac@npac.syr.edu

If you have any comments about this server, send e-mail to webmaster@npac.syr.edu.

Page produced by wwwfoil on Wed Feb 19 1997