Direct Treatment of Uncertaitny

1
High Performance Computational Science/Engineering
- Modeling and Simulation Needs

• NSF CyberInfrastructure Workshop

Gregory J. McRae MIT Chemical Engineering 25th
September 2006
2
Outline of Presentation

• Lessons from Workshops
• Cyberinfrastructure trends
• Future needs
• Conclusions

Key Message This workshop is critical and could
shape the future of the NSF as well as the
national agenda
3
Gregory J. McRae Workshop Experiences

• High Performance Computing
• U.S. DoE Science Advisory Board (Computing)
• Blue Ribbon Panel(s) on HPC
• Pittsburgh Supercomputing Center Scientific
  Advisory Board
• Sandia Management Advisory Board
• NAS/NAE Panels on Computing
• Warren Center, Sydney University
• Queensland Cyber Infrastructure Facility (QCIF)
• NSF Cyberinfrastructure for Combustion
• What have I learned (Apart from saying no)!!

4
Lesson 1- Importance of Computational Science
Report contained lots of examples.
5
Lesson 2- Shaping the Research Agenda
These reports can shape outcomes if a strong case
is made with good, practical and relevant
examples
www.sc.doe.gov/bes/reports/list.html
6
Lesson 3- Regaining the High Ground
John von Neumann B.S. Chemical Engineering, ETH
Zurich
1903-1957
Stored Program Model, ENIAC, Algorithms,
Andrew Grove B.S., Ph.D. Chemical Engineering
Co-founder INTEL,
7
Lesson 4- Impacts From Computational Science
Where are the chemical engineers?
8
Lesson 5- It is not just the hardware, the
solution of problems should drive HPC
9
Energy and Oil Recovery Maximizing Profit
Parameters (porosity, permeability, saturation,
valve settings, market economics,)
Water
Black Oil Model
Oil
Gas
NPV Distributions
10
Problem Characteristics

• Large scale O(104) points, 100s state
  variables
• Nonlinear Chemistry, IPDE operators,
• Wide range of time scales lmax/lmin 1014
• Repeated Solution Exploiting characteristics
• Many uncertainties Parameters, inputs,..
• Decision Making Where to allocate resurces
• High payoffs 1 better yield -gt 1 year supply

11
Effect of New Information on Risk/Value
Conventional Water Injection
Optimized result (Same production rate, 25 extra
recovery)
12
Drill where there are hydrocarbons!!
13
Five Key Trends/Issues

• The widening gap between application performance
  and peak performance of high-end computing
  systems
• Widespread emergence of large, multidisciplinary
  computational science teams in the research and
  applications communities (eScience)
• The flood of scientific data from both
  simulations and experiments being analyzed with
  workflow software
• Network expansion and device collobaration
• Increasing importance to National competitiveness
  scientific, social and economic progress

14
Components of Problem Driven eScience

• Hardware (Computers, sensors,)
• Networks (Hard and wireless)
• Algorithms
• Storage (Hierarchical,)
• Problem Solving Environments
• Impacts not just process
• Support infrastructures
• And
• A vision for the future

Cyber Infrastructure
15
Hardware Technical Trends in Computing

• Moores Law (observation) still going strong
• Smaller, more computing devices every 18 months
• Miniaturization continues
• gt100Gb per square inch hard disk density
• gt128MB memory on a single chip
• Dramatic innovation towards longer battery time
• Low power sensors and CPUs
• Fuel Cell battery (1 month cell phone usage) on
  the horizon
• Smaller, lighter PC, PDA, phone designs enabling
  new networking scenarios
• TVs on Cell phones, Wearable computers, digital
  cash, eBooks

16
Hardware Top500 Supercomputer Sites
LINPACK Benchmarks
www.top500.org
17
Hardware IBM BlueGene/L System 360 TF/s
Steam Iron 50 kWatts/m2
18
Hardware Power Consumption A key Problem
Charles Babbage (1791 1827)
I wish to God these computations had been
executed by steam. Charles Babbage, 1821
19
Hardware -- Hot Chips
20
Hardware Communications Changing
21
Hardware What Mobile Devices can do

• Mobile Devices have 3 capabilities that will have
  broad impact

InitiateActions
Sense theEnvironment
DeliverService
22
Algorithms Needed for Solutions

• State-of-the-art computational science requires
  increasingly diverse and complex
  algorithms
• Only balanced systems that can perform well on a
  variety of problems will meet future scientists
  needs!
• Data-parallel and scalar performance are both
  important

23
Algorithms Science, Algorithms, Needs
24
Algorithms It is not just the hardware
Rational Drug Design
Nanotechnology
Tomographic Reconstruction
Phylogenetic Trees
Biomolecular Dynamics
Neural Networks
Crystallography
Fracture Mechanics
MRI Imaging
Reservoir Modelling
Molecular Modeling
Biosphere/Geosphere
Diffraction Inversion Problems
Distribution Networks
Chemical Dynamics
Atomic Scattering
Electrical Grids
Flow in Porous Media
Pipeline Flows
Data Assimilation
Signal Processing
Condensed Matter Electronic Structure
Plasma Processing
Chemical Reactors
Cloud Physics
Electronic Structure
Boilers
Combustion
Actinide Chemistry
Radiation
Fourier Methods
Graph Theoretic
CVD
Quantum Chemistry
Reaction-Diffusion
Chemical Reactors
Cosmology
Transport
n-body
Basic Algorithms Numerical Methods
Astrophysics
Multiphase Flow
Manufacturing Systems
CFD
Discrete Events
PDE
Weather and Climate
Air Traffic Control
Military Logistics
Structural Mechanics
Seismic Processing
Population Genetics
Monte Carlo
ODE
Multibody Dynamics
Geophysical Fluids
VLSI Design
Transportation Systems
Aerodynamics
Raster Graphics
Economics
Fields
Orbital Mechanics
Nuclear Structure
Ecosystems
QCD
Pattern Matching
Symbolic Processing
Neutron Transport
Economics Models
Genome Processing
Virtual Reality
Astrophysics
Cryptography
Electromagnetics
Computer Vision
Virtual Prototypes
Intelligent Search
Multimedia Collaboration Tools
Computer Algebra
Databases
Magnet Design
Computational Steering
Scientific Visualization
Data Mining
Automated Deduction
Number Theory
CAD
25
Algorithms Multi-scale Engineering
Experimental Data and Quantum Chemistry (Gaussian,
)
Design Optimization (gPROMS, Aspen,)
CFD Models of Reactive Flow (Fluent, StarCD,..)
Seamless Integration/Interoperability of Software
Tools
26
Impacts Speedo and Supercomputers for Swimsuits
Grant Hackett
27
Linkage With Other Disciplines An Example
Identical mathematical structure and common
solution algorithms -- where are advances being
made?
28
Taking Advantage of Problem Knowledge
Levy Process with Jumps (Population Balance Model)
Share Price Volatility
Problem specifi, multi-stage. explicit,
Runge-Kutta Chebyshev
Eigenvalue structure of the discretized problem
Expanded stability domains for time integration
methods
29
Problem Solving Environments (PSE) Usability
The old computing was about what computers could
do, The New Computing is about what people can
do
mitpress.mit.edu/leonardoslaptop
Usable Reliable and comprehensible Universal
Diverse users and varied equipment Useful In
harmony with human needs
30
PSEs eScience Work Environments (QCIF)
31
PSE Make computer usage less frustrating
32
PSE Desirable Characteristics of Environments

• Very fast, accurate, cheap, robust!!
• Be able to treat black box models
• Take advantage of distributed computing
• Ease of use without compromising complexity

33
Networks Connecting to the world
34
Networks A More Generic Topology
Chemical Plant Analog
35
Networks Scientific Workflow and Collaboration
www.neptune.washington.edu
Network of sensors controlled over the internet
36
Networks Universal Access to Knowledge
If there is electricity
you will find Google!
Wikipedia, Google, changing sociological models
37
Networks Fastest Growing Application Ever?
VoIP
38
Infrastructure Cross-Cutting Challenges

• Institutionalize infrastructure
• Broad deployment support at sites
• Software as infrastructure
• Legitimate ( challenging) security concerns
• Expand range of resource sharing modalities
• Research aimed at federating not just data
  computers, but workflow and semantics
• Scale data size, community sizes, etc., etc.
• Reach new application domains
• Sustain current collaboratory pilots, and start
  new ones of similar or greater ambition

39
Hardware Commercial Viability Legacy of Failures

• ACRI French-Italian program
• Alliant Proprietary Crayette
• American Supercomputer
• Ametek
• Applied Dynamics
• Astronautics
• BBN
• CDC gtETA ECL transition
• Cogent
• Convex gt HP
• Cray Computer gt SRC GaAs flaw
• Cray Research gt SGI gt Cray Manage
• Culler-Harris
• Culler Scientific Vapor
• Cydrome VLIW
• Dana/Ardent/Stellar/Stardent
• Denelcor
• Encore
• Elexsi

• Goodyear Aerospace MPP SIMD
• Gould NPL
• Guiltech
• Intel Scientific Computers
• International Parallel Machines
• Kendall Square Research
• Key Computer Laboratories searching again
• MasPar
• Meiko
• Multiflow
• Myrias
• Numerix
• Pixar
• Parsytec
• nCube
• Prisma
• Pyramid Early RISC
• Ridge
• Saxpy

40
A Strategy for the Future

• A vision to Predict with confidence
• Productivity of physics and engineering users is
  vital. Its the problems to be solved, not the
  computers
• A needs-based strategy based on increasing
  high-end computing capabilities
• No successful program is an island success
  depends on success of the community
• Remain agile and ready to capitalize on
  scientific breakthroughs

41
Contact Information
Gregory J. McRae Department of Chemical
Engineering Room 66-372 Massachusetts Institute
of Technology Cambridge, MA 02139 (617) 253
6564 [phone] (fax) mcrae_at_mit.edu
(email) http//www.mit.edu/cheme