Earth System Science Challenges for the NextGeneration GIS Alan M. Gaines Division of Earth Sciences

1
National Science Foundation
Arlington, VA http//www.nsf.gov
2
Earth System CollaboratoryA Concept Paper

• Background
• Earth System Collaboratory Concept
• Resources Needed
• Computational
• Financial
• Off Shore Activities
• Opportunity for Funding?

3
BACKGROUND

• Distributed Climate Simulation Laboratory
• Opportunity fund award of 2 M (SDSC, PSC,
  NCAR)
• Proposed candidate for MRE 1995 - 2000
• NSF-DOE SSI
• PITAC

4
Distributed Climate Simulation Laboratory

• SDSC and NCAR

5
Schematic Scenario of Collaboration between GCRP
and HPCC
Regional chemistry model run with boundary
conditions from NCAR climate model
vBNS 155 Mbits/sec
CSC
NCSA
PSC
NCAR
Hydrological cycle simulation on scale parallel
architecture computer
SDSC
Visualization Distributed Cluster Computing
Linkages to others e.g. Integrated Assessment
Communities
CLS MSS
Other Models run on Paragon e.g. Ocean
6
MAJOR RESEARCH EQUIPMENTACCOUNT

• Proposed 1995 - 2000

7
Major Research Equipment Efforts
8
The NSF/DOE Teraflop Computing Program

• Collaborative Effort

9
Background
First (Fall 97) there was ACPI -- the DOE
Accelerated Climate Prediction Initiative Then
(Spring 98) there was SSI -- the DOE Strategic
Simulation Initiative Finally (Summer 98) NSF
joined the fray
10
Background

• Workshop held at NAS on July 30-31
• Convened by DOE and NSF
• 200 people attended
• Goal National network of Teraflop
  supercomputers for civilian use (ASCI
  counterpart)
• Applications include climate, combustion, and
  biotechnology
• Other potential participants include NIH, NOAA,
  and NASA
• Total funding 1B over 4 years
• First funding in FY00? New money?
• Computer Experts urge new Federal
  initiatives Science, 7 Aug. 98, pg. 762

11
President's Information TechnologyAdvisory
CommitteeInterim Report to the President

• August 1998

12
Complimentary Activity

• Presidents Information Technology Advisory
  Committee (PITAC)
• 26 members from industry and academe
• Concerned about support for long-term, basic
  research in Information Technology (IT)
• Recommend additional 1B over 5 years
• NSF may be the lead agency
• Primary focus is IT, but recommendations include
  additional support for research on high-end
  computing
• 1. Report urges U.S. to take the Long View,
  Science, 21 Aug. 98, pg.. 1125
• 2. See also www.ccic.gov/ac/interim/

13
EARTH SYSTEM COLLABORATORY CONCEPT
14
EARTH SYSTEM COMPONENTS
14
15
EARTH SYSTEM COLLABORATORYGOAL

• Greatly accelerate the rate of scientific
  progress in understanding the Earth System in
  support of both research and assessment.
• Establish an integrated core of 5-7 major
  intellectual centers to enable scientists have
  easy-to-use, powerful facilities for
  collaboration.

16
EARTH SYSTEM COLLABORATORYSYSTEM

• Fuses technology assets together to facilitate
  cross-center researcher enabling a vast reduction
  of the traditional barriers of time and
  geographic separation.
• Gigahertz desktop systems and Earth System
  Situation Rooms at each center will present
  Collaboratory resources over high-bandwidth
  networks using a new generation of Web-enabled
  technologies.

17
EARTH SYSTEM COLLABORATORYEARTH SYSTEM SITUATION
ROOMS

• Earth System Situation Rooms at each center will
  provide a sophisticated physical nexus for
  scientific review and presentation.
• Groups of researchers will be able to
  interactively explore and study model results,
  compare historical archives and other findings,
  and interact with researchers at other
  Collaboratory situation rooms as well as
  geographically dispersed participants with
  commodity desktop PCs.
• These rooms will enable virtual workshops and
  conferences.
• Each situation room will combine large-format
  display systems, advanced visualization
  capabilities, and virtual conferencing.

18
(No Transcript)
19
(No Transcript)
20
EARTH SYSTEM COLLABORATORYDATA

• Central to Collaboratory will be a massive
  collection of simulated and observational data
  holdings.
• The Collaboration System will present both online
  and nearline data holdings as common
  Web-accessible objects with powerful search and
  indexing mechanisms layered on top.
• These data holdings will be seamlessly accessible
  for both simulation as well as post-simulation
  analysis.

21
EARTH SYSTEM COLLABORATORYVISUALIZATION AND DATA
ANALYSIS

• The Collaboratory will enable teraflop
  computation and terabyte data archival
  capability.
• Capability creates challenge How does a human
  analyze, in a timely manner, the staggering
  quantities of data that a single hour of
  computation could produce?
• Will technology provide amalgam of visualization,
  data analysis tools and scale networks sufficient
  to gain knowledge from information?

22
HOW MUCH COMPUTATIONAL INFRASTRUCTURE DO WE NEED
HOW MUCH WILL IT COST?
23
Weather and Climate Prediction Requires a Large
Increase in Computing Power by 2003

• Scale
  Resolution
• Ocean 10 km
• Atmosphere 30 km
• Land surface 1 km
• Processes
• Cloud-radiation interaction
• Biogeochemistry, hydrology, river runoff
• Aerosols and greenhouse gases
• Surface-atmosphere exchange
• Detailed mountains

24
Requirements for Advanced Computational
Infrastructure and Environments

• Advances in storage, analysis and visualization
  of petabyte data sets
• Multi-institutional problem solving environment
• Regional collaborators for assessment and impacts
  research
• Better numerical and computational methods for
  parallel systems
• Breakthroughs in scaling for parallel systems

25
DOE/NSF Accelerated Program
multiple ensemble forecasts
Accelerated Program
1000
high resolution ensemble forecasts
100
coupled model validation
TERAFLOPS
model component validation
10
Business as Usual
1
0.1
1999
2003
26
(No Transcript)
27
Emerging Computational Opportunities (Requiremen
ts for Solution in 1 Week)
Recent Computations by NSF Grand Challenge
Research Teams
Next Step Projections by NSF Grand Challenge
Research Teams
Long Range Projections from Recent Applications
Workshop
1014
Turbulent Convection in Stars
ASCI in 2004
NSF in 2004 (Projected)
1012
MEMORY
QCD
1999 NSF Leading Edge
BYTES
Computational Cosmology
100 year climate model in hours
1010
1995 NSF Capability
Atomic/Diatomic Interaction
Molecular Dynamics for Biological Molecules
108
1012
1014
1016
1018
1020
108
1010
MACHINE REQUIREMENT IN FLOPS
28
EARTH SYSTEM COLLABORATORYCOST

• In July, 1998 NSF/Geosciences asked NCAR to
  develop a plan spanning FY00-05 that would
  culminate with 40-50 Teraflops sustained capacity
  by end of FY05.
• The team
• Bill Buzbee Tom Angel
• Pete Peterson Marla Meehl
• Bernie OLear Don Middleton
• Steve Hammond Aaron Andersen

29
Moores Law and U.S. Microprocessor DSMs
ASSUMPTION
30
Moores Law and U.S. Vector Systems (SV1)
ASSUMPTION
31
THE TECHNOLOGICAL INFRASTRUCTURE REQUIRED FOR
EARTH SYSTEM COLLABORATORY

• Collaboration system
• High Speed Communication system
• Distributed visualization systems
• Distributed computing
• Distributed storage
• Operational infrastructure
• Space for equipment and staff
• Power and cooling for equipment
• Systems and networking staff
• Expertise in all aspects of parallel and
  distributed computing, visualization, etc.

32
NSF Earth System Collaboratory Annual Cost
FISCAL YEAR 2000
33
Estimated Costs in Millions of Dollars
34
Are Partnerships Possible?

• DOE, NASA, NOAA, NIH?

35
A Potential NSF/DOE Partnership
NSF Earth System Collaboratory
Initiative GEO CISE
DOE SSI Climate (ACPI) Combustion Basic
Research
NCAR / PACI
LANL
LLNL
? a) A triad of Teraflop systems to support Earth
System modeling b) NSF is an equal partner
36
Maintaining Intellectual Leadership in Earth
System Science

• What is going on off-shore.

37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
What are the prospect of receiving support for
the Earth System Collaboratory?
41
PROS

• A unifying theme to accelerate interdisciplinary
  research
• Could involve all Directorates at NSF
• Would require research activities that are
  similar to those recommended in PITAC
• Builds on successful model of a Collaboratory
  funded by GEO MPS and complements and enhances
  activities underway within NPACI
• Supports the objectives of KDI
• Network for Earthquake Engineering Simulation
  request for MRE funding sets a precedent.

42
CONS

• Requested funding is about 1.44 Billion over 5
  years
• Outside the guidelines set by OMB
• New paradigm for NSF
• Top priority of NSF is likely to be our response
  to PITAC recommendations
• Partnership between NPACI Centers and NCAR not
  sufficiently developed
• Partnership between NSF and DOE with respect to
  SSI remains unclear

43
COMMENTS QUESTIONS