WTEC International Assessment of Simulation-Based Engineering and Science: Education

1
WTEC International Assessment of
Simulation-Based Engineering and Science
Education

• Celeste Sagui
• North Carolina State University
• Sharon Glotzer
• University of Michigan

Sponsors NSF, DOE, DOD, NIH, NASA, NIST
2
Simulation-Based Engineering Science

• SBES involves the use of computer modeling and
  simulation to solve mathematical formulations of
  physical models of engineered and natural systems
• SBES or computational science engineering
  is an established (though not mature) field.

3
SBES Why now?

• A tipping point in SBES
• Computer simulation is more pervasive today, and
  having more impact, than ever before - hardly a
  field untouched
• Fields are being transformed by simulation
• Reached a useful level of predictiveness
  complements traditional pillars of science

4
SBES Why now?

• A tipping point in SBES
• Computers are now affordable and accessible to
  researchers in every country around the world.
• The near-zero entry-level cost to perform a
  computer simulation means that anyone can
  practice SBES, and from anywhere.
• Flattening world of computer simulation that
  will continue to flatten - everyone can do it.

5
SBES Why now?

• A tipping point in SBES
• US, Japanese, EU companies are building the next
  generation of computer architectures, with the
  promise of thousand-fold or more increases of
  computer power coming in the next half-decade.
• These new massively multicore computer chip
  architectures will allow unprecedented accuracy
  and resolution, as well as the ability to solve
  the highly complex problems that face society
  today.

6
SBES Why now?

• A tipping point in SBES
• The toughest scientific and technological
  problems facing society today are big problems
• alternative energy sources and global warming
• sustainable infrastructures
• mechanisms of life, curing disease and
  personalizing medicine.
• These problems are complex and messy, and their
  solution requires a partnership among experiment,
  theory and simulation, and among industry,
  academia and government, working across
  disciplines.

7
SBES Why now?

• Simulation is key to scientific discovery and
  engineering innovation.
• Recent reports argue the United States is at risk
  at losing of its competitive edge.
• Our continued capability as a nation to lead in
  simulation-based discovery and innovation is key
  to our ability to compete in the 21st century.

8
Previous SBES study

• Our study builds upon previous efforts
• Workshops run by NSF Engineering Directorate
• NSF Blue Ribbon Panel report chaired by J.
  Tinsley Oden, May 2006 - lays out intellectual
  arguments for SBES
• SBES broadened to SBES
• many previous reports on computational science

http//www.nsf.gov/pubs/reports/sbes_final_report.
pdf
9
SBES - A National Priority

• The Promise Advances in mathematical modeling,
  in computational algorithms, in the speed of
  computers, and in the science and technology of
  data intensive computing, have brought the field
  of computer simulation to the threshold of a new
  era, an era in which unprecedented improvements
  in the health, security, productivity, and
  competitiveness of our nation may be possible. A
  host of critical technologies are on the horizon
  that cannot be understood, developed, or utilized
  without simulation methods.

From Oden report
10
WTEC SBES Study Sponsors

• To inform program managers in U.S. research
  agencies and decision makers of the status,
  trends and activity levels in SBES research
  abroad, these agencies sponsored this study
• National Science Foundation (NSF)
• Department of Energy
• Department of Defense
• National Institutes of Health
• NASA
• National Institute of Standards and Technology

11
Overall Scope Objectives of WTEC International
Study

• Study designed to
• Gather information on the worldwide status and
  trends of SBES research
• State of the art, regional levels of activities
• US leadership status
• Opportunities for US leadership
• Disseminate this information to government
  decision makers and the research community
• Findings, not recommendations

12
Structure of Study

• Primary thematic areas
• Life sciences and medicine
• Materials
• Energy and sustainability
• Core cross-cutting issues
• Next-generation algorithms and high performance
  computing
• Multiscale simulation
• Simulation software
• Validation, verification, and quantifying
  uncertainty
• Engineering systems
• Big data and data-driven simulations
• Education and training
• Funding

13
The SBES Study Team

• Panelists

• Advisors

• Sharon Glotzer (Chair), U Michigan
• Sangtae Kim, NAE (Vice-chair), Purdue
• Peter Cummings, Vanderbilt/ORNL
• Abhi Deshmukh, Texas AM
• Martin Head-Gordon, UC Berkeley
• George Karniadakis, Brown U
• Linda Petzold, (NAE) UC Santa Barbara
• Celeste Sagui, NC State U
• Matsunoba Shinozuko, (NAE) UC Irvine

• Tomas de la Rubia, LLNL
• Jack Dongarra, (NAE) UTK/ORNL
• James Duderstadt (NAE), U Michigan
• David Shaw, D.E. Shaw Research
• Gilbert Omenn (IOM), U Michigan
• J. Tinsley Oden (NAE), UT Austin
• Marty Wortman, Texas AM

14
Study Process Timeline

• US Baseline Workshop November 2007
• Bibliometrics analysis
• Panel visited 57 sites in Europe, Asia
• Universities, national labs, industrial labs
• Also conversations, reports, research papers,
  bibliometric analysis provided basis for
  assessment
• Public workshop on study findings in April 2008
• Final report now in review
• Research directions planning workshop in April
  2009

15
Sites Visited in China December 2007
Peking Univ./CCSE, Tsinghua Univ./DEM, ICCAS,
ICMSEC/CAS, IPE/CAS,
Dalian Univ. of Technology
SSC, Shanghai Univ., Fudan Univ.
http//www.lonelyplanet.com/maps/asia/china/
16
Sites Visited in Japan December 2007
RIKEN/ACCC
NIMS/CMSC, RICS/AIST
Kyoto Univ.
CRIEPI, SBI, Univ. Tokyo
Japan Agency for Marine-Earth ST (ESC), Nissan
Research Center, Mitsubishi Chemicals
Toyota Central RD Labs., IMS
http//www.ease.com/randyj/japanmap.htm
17
Sites Visited in Europe February 2008
Unilever RD, Daresbury Lab
Univ. Oxford, Univ. Cambridge, Unilever Centre
Univ. College London, The Thomas Young Centre
Vrije Univ.
DTU
ZIB
IWM,BASF, ITTPE, Univ. Karlsruhe
IFP
Paris Simulation Network
Tech. Univ. Munich
IMFT, ENSEEIHT, IRIT
CERN, EPFL/IACS, ETH, IBM, Univ. Zürich
Eni SpA, MOX Remote site visit
CIMNE, ICMAB/CSIC
57 sites/36 in Europe
http//www.europeetravel.com/maps/western-europe-m
ap.htm
18
Major Trends in SBES Research
www.wtec.org/sbes
C. Sagui and S.C. Glotzer
19
Life sciences medicine, materials, and energy
sustainability are among most likely sectors to
be transformed by SBES

• SBES is changing the way disease is treated, the
  way surgery is performed and patients are
  rehabilitated, the way we understand the brain
• SBES is changing the way materials components
  are designed, developed, and used in all
  industrial sectors
• E.g. ICME (National Academies Report 2008, T.
  Pollock, et al)
• SBES is aiding in the recovery of untapped oil,
  the discovery utilization of new energy
  sources, and the way we design sustainable
  infrastructures

20
Findings Top FourMajor Trends in SBES Research

1. Data-intensive applications (esp Switzerland and
   Japan)
2. Integration of (real-time) experimental and
   observational data with modeling and simulation
   to expedite discovery and engineering solutions
3. Millisecond timescales for proteins and other
   complex matter with molecular resolution
4. Science-based engineering simulations (US slight
   lead)
5. Increased fidelity through inclusion of physics
   and chemistry
6. Multicore for petascale and beyond not just
   faster time to solution - increased problem
   complexity
7. Cheap GPUs today give up to 200x speed up on
   hundreds of apps!

21
Threats to United States Leadership in SBES
www.wtec.org/sbes
C. Sagui and S.C. Glotzer
22
Some general trends RD map
Figure from the council on Competitiveness
Competitiveness Index Where America stands
(2007). Data from Main Science and Technology
Indicators 2006 (OECD 2006).
23
US share of global output in ST
New SE PhDs
Figure from the council on Competitiveness
Competitiveness Index Where America stands
(2007). Data from Main Science and Technology
Indicators 2006 (OECD 2006) NSFs Science and
Engineering Indicators (NSB 2006), and the U.S.
Patent and Trademark Office.
24
Threats to US leadership in SBESEducation
Impacts

• Finding 1 The world of computing is flat, and
  anyone can do it. We must do it better, and
  exploit new architectures before those
  architectures become ubiquitous ? crucial to
  train next generations of SB engineers and
  scientists.

25
Threats to US leadership in SBES

• Top 500 list US at top today. But Japan, France,
  Germany have world-class resources, faculty and
  students and are committed to HPC/SBES for long
  haul.
• Japan has an industry-university-govt roadmap out
  to 2025 (exascale)
• Germany investing nearly US1B in new HPC push,
  also with EU
• Cheap to start up, hire in SBES (e.g. India)
• 100M NVIDIA GPUs w/CUDA compilers worldwide
• Every desktop, laptop, etc. with NVIDIA card in
  last two years
• Speed-ups of factors up to 1000. Applications
  from every sector.

26
Threats to US leadership in SBESEducation
Impacts

• Finding 2 A persistent pattern of subcritical
  funding overall for SBES threatens US leadership
  and continued needed advances amidst a recent
  surge of strategic investments in SBES abroad.
  The surge reflects recognition by those countries
  of the role of simulations in advancing national
  competitiveness and its effectiveness as a
  mechanism for economic stimulus.

27
Threats to US leadership in SBES

• Germany restructuring universities new
  univ-industry partnerships
• 20 year-on-year increase effective
  restructuring to support collaboration
• E.g. Fraunhofer IWM Karlsruhe University
  (16M/yr, 44 industry, 50 SBES)
• Japan committed to HPC, and leads US in bridging
  physical systems modeling to social-scale
  engineered systems
• Singapore and Saudi Arabia - in SE
• Expect increased China and India presence in
  scientific simulation software RD and SBES
  generally over next decade due to new academic
  industry commitment, new government

28
Threats to US leadership in SBES

• China not yet a strong US competitor, but SBES
  footprint changing rapidly
• China contributes 13 of the worlds output in
  simulation papers, second to US at 27 and
  growing (but publish in lt1st tier journals and
  cited less)
• Non-uniform quality overall, but many high
  quality examples
• Strategic change towards innovation, and
  recognition by industry and State that innovation
  requires simulation
• Chinas ST budget has doubled every 5 years
  since 1990
• 70 to top 100 universities (80 all PhDs, 70
  all , 50 all international,30 all UGs)
• Recognition of need to train new generation of
  computationally-savvy students, and new State
  to do this under new VM of Education
• gt211 Fund US1B/year, all projects must have
  integrated simulation component

29
Threats to US leadership in SBES

• We found healthy levels of SBES funding for
  company-internal projects, underscoring
  industrys recognition of the cost-effectiveness
  and timeliness of SBES research.
• The mismatch vis a vis the public-sectors
  investment level in SBES hinders workforce
  development.
• We saw many examples of companies (including US
  auto and chemical companies) working with EU
  groups rather than US groups for better IP
  agreements.

30
Drivers and barriers for HPC usage in
industryUS Council on Competitiveness Report,
2008

• Hurdles There are three systemic barriers to
  HPC 1) Lack of application software, 2) access
  to talent, 3) Cost constraints (capital,
  software, expertise).
• Most of firms revealed they have important
  problems they can not solve on their desktop
  systems. Over 60 of firms would be willing to
  pay outside organizations (non-profits,
  engineering services companies, or major
  universities) for realizing the benefits of HPC.
• The survey implications are sobering critical
  U.S. supply chains and the leadership of many
  U.S. industries may be at risk if more companies
  do not embrace modeling and simulation with HPC.

31
Threats to US leadership in SBES

• Because SBES is often viewed within the US more
  as an enabling technology for other disciplines,
  rather than a discipline in its own right,
  investment in and support of SBES is often not
  prioritized as it should be at all levels of the
  RD enterprise.
• We found that investment in computational science
  in the US and the preparation of the next
  generation of computational researchers remains
  insufficient to fully leverage the power of
  computation for solving the biggest problems that
  face the US going forward.

32
Threats to US leadership in SBES

• Finding 3 Inadequate education and training of
  the next generation of computational scientists
  threatens global as well as US growth of SBES.
  This is particularly urgent for the US, since
  such a small percentage of its youths go into SE.

33
Threats to US leadership in SBES

• Finding 2 Inadequate education and training of
  the next generation of computational scientists
  threatens global as well as US growth of SBES.
  This is particularly urgent for the US, since
  unless we prepare these researchers to use the
  next generation of computer architectures we are
  developing, we will not be able to exploit their
  game-changing capabilities.

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
34
Education and Training some statistics

• US has most citations and top-cited publications
  but EU has surpassed in number of articles

SE articles and citations in all fields. From
Science and Engineering Indicators 2008 (NSB
2008).
35
Education and Training some statistics

• US has been surpassed in number of PhDs in SE

Number of PhDs earned in Europe, Asia and North
America (2004). From Science and Engineering
Indicators 2008 (NSB 2008).
36
Education and Training some statistics

• First-time, full-time graduate enrollment in SE

Dotted foreigners Solid permanent residents
NSF (2007)
37
Education and Training some statistics

• Left Foreign students enrolled in tertiary
  education, 2004. Right SE doctoral degrees
  earned by foreign students

SE Indicators (2008)
38
Education and Training some statistics

• Academic RD share of all RD, for selected
  countries (SE Indicators, 2008)

39
Education and Training some statistics

• Natural Sciences and Engineering degrees per
  hundred 24-year olds, by country (SE
  Indicators, 2008)

40
Education and Training some statistics

• SE postdoctoral students at US universities, by
  citizenship (SE Indicators, 2008)
• Percentage of visa post-docs
• -Biological Sciences 59
• -Computer Sciences 60
• -Engineering 66
• -Physical Sciences 64

41
Education and TrainingKey Findings
www.wtec.org/sbes
C. Sagui and S.C. Glotzer
42
Education and TrainingKey findings

• Finding 1 There is increasing Asian and European
  leadership in SBES education due to dedicated
  funding allocation and industrial participation.

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
43
Finding 1 (a) Increasing Asian leadership due to
funding allocation and industrial participation
in education

• Japan committed to HPC, and leads US in bridging
  physical systems modeling to social-scale
  engineered systems
• Japan Earth Simulation Center (Life Simulation
  Center) developing new algorithms, specially
  multiscale and multiphysics. Govt investing in
  software innovation in algorithms will drive
  hardware.
• Systems Biology Institute (Japan) funded by
  Japanese government for 10 years. Software
  infrastructure Systems Biology Markup Language
  (SBML), Systems Biology Graphical Notation
  (SBGN), CellDesigner, and Web 2.0 Biology.
  Difficult to publish software, the merit system
  in this lab values software contributions as well
  as publications.
• University of Tokyo 21st Century Center of
  Excellence (COE) Program 28 worldclass research
  and education center in Japanese Universities ?
  Global COE
• Singapore and Saudi Arabia in SE (KAUST
  university, with 80B endowment)
• Increased China and India presence in scientific
  simulation software RD and SBES over next
  decade due to new academic industry commitment,
  new government
• Institute of Process Engineering (P.R. China)
  50 of research funding comes from industry
  (domestic and international significant funding
  from the petro-chemical industry). Significant
  government funding through the National Natural
  Science Foundation of China and the Ministry of
  Science and Technology (main focus multiscale
  simulations for multiphase reactors ).
• Tsinghua University Department of Engineering
  Mechanics Strong interaction of RD centers with
  industry and multinational companies.
• Fudan University, Shanghai strong emphasis on
  education, first analytical work then
  computational. Prof. Yang is director of leading
  computational polymer physics group and Vice
  Minister of Education has allocated funding for
  SBES and for 2000 students/year to study abroad.

44
Finding 1 (a) Increasing Asian leadership due to
funding allocation and industrial participation
in education

• China not yet a strong US competitor, but SBES
  footprint changing rapidly
• China contributes 13 of the worlds output in
  simulation papers, second to US at 27 and
  growing (but publish in lt1st tier journals and
  cited less)
• Non-uniform quality overall, but many high
  quality examples
• Strategic change towards innovation, and
  recognition by industry and State that innovation
  requires simulation
• Chinas ST budget has doubled every 5 years
  since 1990
• 70 to top 100 universities (80 all PhDs, 70
  all , 50 all international,30 all UGs)
• Recognition of need to train new generation of
  computationally-savvy students, and new State
  to do this under new VM of Education
• gt211 Fund US1B/year, all projects must have
  integrated simulation component

45
Finding 1 (b) Increasing European leadership due
to funding allocation and industrial
participation in education

• Center for Biological Sequence Analysis
  (Bio-Centrum-DTU, Denmark) Danish Research
  Foundation, the Danish Center for Scientific
  Computing, the Villum Kann Rasmussen Foundation
  and the Novo Nordisk Foundation (US100M), other
  institutions in European Union, industry and the
  American NIH (bioinformatics, systems biology).
• CIMNE International Center for Numerical
  Methods in Engineering (Barcelona, Spain)
  independent research center, now as a consortium
  between Polytechnic University of Catalonia, the
  government of Catalonia, and the federal
  government annual funding 10M from external
  sources, focused on SBES research, training
  activities and technology transfer.
• Germany restructuring universities new
  univ-industry partnerships. German research
  foundation (DFG) has provided support for
  collaborative research centers (SBF), transregion
  projects (TR), transfer units (TBF), research
  units (FOR), Priority programs, and Excellence
  Initiatives. Many of these are based on or have
  major components in SBES (Stuttgart, Karlsruhe,
  Munich) and strong connections with industry
• Fraunhofer Institute for the Mechanics of
  Materials (Germany) 15.5M/year, 44 from
  industry and 25-30 from government. Significant
  growth recently (10 per year). Fully 50 of
  funding goes to SBES (up from 30 5 years ago)
  (applied materials modeling), 50,000 euro
  projects awarded to PhDs to work in the institute
  in topic of their choice.

46
Finding 1 (b) Increasing European leadership due
to funding allocation and industrial
participation in education

• Partnership for Advanced Computing in Europe
  (PRACE) coalition of 15 countries led by
  Germany and France, based on the infrastructure
  roadmap outlined in the 2006 report of the
  European Strategy Forum for Research
  Infrastructures. This roadmap aims to install
  five petascale systems around Europe beginning in
  2009, in addition to national HPC facilities and
  regional centers.
• TALOS Industry-govmt alliance to accelerate the
  development in Europe of new-generation HPC
  solutions for large-scale computing systems.
• DEISA consortium of 11 leading European national
  supercomputing centers to operate a
  continent-wide distributed supercomputing
  network, similar to TeraGrid in the United States.

C. Sagui and S.C. Glotzer
SIAM, CSE09
47
Finding 2 New centers and programs for education
and training in SBES all of interdisciplinary
nature

• CBS (BioCentrum-DTU) MSc in Systems Biology and
  in Bioinformatics loosely structured, not linked
  to any department in particular. Real-time
  internet training (all lectures, exercises and
  exams), with typically 5050 students
  onsiteinternet. International exchange highly
  encouraged, students can take their salary and
  move anywhere in the globe for half a year.
• CIMNE (Barcelona) main especiality is courses
  and seminars on the theory and application of
  numerical methods in engineering. In last 20
  years, CIMNE has organized 100 courses, 300
  seminars, 80 national and international
  conferences, published 101 books, 15 educational
  software 100s of research and technical reports
  and journal papers.
• ETH Zurich pioneering CSE program (MSc and BSc)
  combining several departments, successful with
  grads and postdocs taking the senior level
  course.
• Technical University of Munich and Leibnitz
  Supercomputing Center Many CSE programs (i)
  BGCE, a Bavaria-wide MSc honors program (ii)
  IGSSE postgraduate school (iii) Center for
  Simulation Technology in Engineering (iv)
  Centre for Computational and Visual Data
  exploration (v) International CSE Master program
  multidisciplinary involving 7 departments also
  allows for industrial internship (iv) Software
  project promotes development of software for
  HPC/CSE as an educational goal (v) many, many
  other programs with other universities and
  industry.

48
Finding 3 EU and Asian Centers are attracting
more international students from all over the
world (including US)

• Japan International Center for Young Scientists
  (Comp. Mat. Science Center Nat. Inst. Mat.
  Sc.) English, interdisciplinary, independent
  research, high salary, research grant support (5M
  yen/year). COE aimed at attracting international
  students. Below 120,000 international students
  enrolled in Japanese universities, PM wants to
  increase number to 300,000.
• China 211 and 985 programs to build
  world-class universities. 200,000 international
  students from 188 countries came in 2007. Main
  donors Korea, Japan, US, Vietnam, Thailand
• King Abdullah University of Science and
  Technology (KAUST) recruiting computational
  scientists and engineers at all levels,
  attracting best and brightest from Middle East,
  India and China.
• Australia targeting Malaysia and Taiwan

49
Finding 3 EU and Asian Centers are attracting
more international students from all over the
world (including US)

• CBS (BioCentrum-DTU) The internet courses are
  used to attract international students (cost 20
  more effort but bring lots of money, always
  oversubscribed).
• CIMNE (Barcelona) (i) introduced an
  international course for masters in computational
  mechanics for non-European students. This is 1st
  year with 30 students. Four universities involved
  in this course (Barcelona, Stuttgart, Swansea and
  Nantes). (ii) Web environment for distance
  learning, also hosting a Master Course in
  Numerical methods in Engineering and other
  postgraduate courses. (iii) the classrooms
  physical spaces for cooperation in education,
  research and technology located in Barcelona,
  Spain, Mexico, Argentina, Colombia, Cuba, Chile,
  Brazil, Venezuela and Iran.
• ETH Zurich number of international students has
  increased dramatically (Asian, Russian).
• Vrije University Amsterdam 50 graduate students
  come from outside the Netherlands (mainly Eastern
  Europe).
• LRZ in TUM Munich 80 SBES students in MSc
  programs come from abroad Near East, Asia,
  Eastern Europe, Central and South America.
• United Kingdom ranks 2nd in world (after US) in
  attracting international students
• Spain, Germany and Italy among others are
  capturing more and more of the latin american
  student market, which has shifted its traditional
  preference for the US in favor of Europe.

50
Finding 4 Pitfall of interdisciplinary
education breadth vs depth

• Educational breadth comes at the expense of
  educational depth. e.g., in ETH Zurich the CSE
  faculty choose physics or chemistry students when
  dealing with research issues and CS majors for
  software development. General feeling that CSE
  students can spend too much time on the format
  of the program, without really thinking the
  underlying science beneath.
• To solve grand challenges in a field, solid
  knowledge of core discipline is crucial.
• Appropriate evaluation of scientific performance
  difficult to come up with credit assignation in
  an interdisciplinary endeavor. Also, hidden
  innovation phenomena (who gets credit when code
  is run by other than author).

51
Finding 5 Demand exceeds supply academia vs
industry

• Huge demand for qualified SBES students who get
  hired immediately after MSc, dont go into PhDs.
  Good to maintain a dynamical market force but
  academia would like to see more students that
  continue a tradition of free research.
• Pharmaceutical, chemical, oil, (micro)electronics,
  IT, communications, software companies
  automotive and aerospace engineering finance,
  insurance, environmental institutions, etc.

52
Finding 6 Inadequate education training
threatens global advances in SBES a
worldwide concern

• Insufficient exposure to computational science
  engineering and underlying core subjects at high
  school and undergraduate level, particularly in
  the US
• Increased topical specialization beginning with
  graduate school
• Insufficient training in HPC an educational
  gap
• Gap b/t domain science courses and CS courses
  insufficient continued learning opportunities
  related to programming for performance
• Most students use codes as black boxes who will
  be innovators?
• Exception pockets of excellence, ie, TUM,
  Stuttgart, Karlsruhe
• No real training in software engineering for
  sustainable codes
• Little training in Uncertainty Quantification,
  Validation Verification, risk assessment
  decision making

53
Education and Training are crucialNext-generatio
n Architectures and Algorithms

• Finding 1 The many orders-of-magnitude in
  speedup required to make significant progress in
  many disciplines will come from a combination of
  synergistic advances in hardware, algorithms, and
  software, and thus investment and progress in one
  will not pay off without concomitant investments
  in the other two.
• Finding 2 The US leads both in computer
  architectures (multicores, special-purpose
  processors, interconnects) and applied algorithms
  (e.g., ScaLAPACK, PETSC), but aggressive new
  initiatives around the world may undermine this
  position.
• At present the EU leads the US in theoretical
  algorithm development.
• Finding 3 The US leads in the development of
  next-generation supercomputers, but Japan,
  Germany committed, and China now investing in
  supercomputing infrastructure.

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
54
Education and Training are crucialScientific
and Engineering software developments

• Finding 1 Around the world, SBES relies on
  leading edge (supercomputer class) software used
  for the most challenging HPC applications,
  mid-range computing used by most scientists and
  engineers, and everything in between.
• Finding 2 Software development leadership in
  many SBES disciplines remains largely in US
  hands, but in an increasing number of areas it
  has passed to foreign rivals, with Europe being
  particularly resurgent in software for mid-range
  computing, and Japan particularly strong on
  high-end supercomputer applications. In some
  cases, this leaves the US without access to
  critical scientific software.
• Finding 3 The greatest threats to US leadership
  in SBES come from the lack of reward,
  recognition and support concomitant with the long
  development times and modest numbers of
  publications that go hand-in-hand with software
  development the steady erosion of support for
  first rate, excellence-based single investigator
  or small-group research in the US and the
  inadequate training of todays computational
  science and engineering students the would-be
  scientific software developers of tomorrow..

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
55
Opportunities for the US to gain or reinforce
lead in SBES

• Finding 1 There are clear and urgent
  opportunities for industry-driven partnerships
  with universities and national laboratories to
  hardwire scientific discovery to engineering
  innovation through SBES.
• This would lead to new and better products, as
  well as development savings both financially and
  in terms of time.
• National Academies report on Integrated
  Computational Materials Engineering (ICME), which
  found a reduction in development time from 10-20
  yrs to 2-3 yrs with a concomitant return on
  investment of 31 to 91.

www.wtec.org/sbes
56
Opportunities for the US to gain or reinforce
lead in SBES

• Finding 2 There is a clear and urgent
  opportunity for new mechanisms for supporting
  SBES RD.
• Support and reward for long-term development of
  algorithms, middleware, software, code
  maintenance and interoperability.
• Although scientific advances achieved through the
  use of a large complex code is highly lauded,
  the development of the code itself often goes
  unrewarded.
• Community code development projects are much
  stronger within the EU than the US, with national
  strategies and long-term support.
• investment in math, software, middleware
  development always lags behind investment in
  hardware

www.wtec.org/sbes
57
Opportunities for the US to gain or reinforce
lead in SBES

• Finding 3 There is a clear and urgent
  opportunity for a new, modern approach to
  educating and training the next generation of
  researchers in high performance computing for
  scientific discovery and engineering innovation.
• Must teach fundamentals, tools, programming for
  performance, verification and validation,
  uncertainty quantification, risk analysis and
  decision making, and programming the next
  generation of massively multicore architectures.
  Also, students must gain deep knowledge of their
  core discipline.

58
For more information and final reportwww.wtec.or
g/sbes
59
(No Transcript)
60
Education and Training WTEC bibliometrics study

• The growth of number of publications in SBES
  worldwide is double the number of all SE
  publications (5 vs 2.5).
• In 2007, US dominated the world SBES output at
  27, but China moved 2nd place at (13).
• EUR-12 have larger SBES output than US, with
  difference increasing over time.

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
61
Threats to US leadership in SBES

• We found healthy levels of SBES funding for
  company-internal projects, underscoring
  industrys recognition of the cost-effectiveness
  and timeliness of SBES research.
• The mismatch vis a vis the public-sectors
  investment level in SBES hinders workforce
  development.
• We saw many examples of companies (including US
  auto and chemical companies) working with EU
  groups rather than US groups for better IP
  agreements.

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
62
Drivers and barriers for HPC usage in
industryUS Council on Competitiveness Report,
2008

• Hurdles There are three systemic barriers to
  HPC 1) Lack of application software, 2) access
  to talent, 3) Cost constraints (capital,
  software, expertise).
• Most of firms revealed they have important
  problems they can not solve on their desktop
  systems. Over 60 of firms would be willing to
  pay outside organizations (non-profits,
  engineering services companies, or major
  universities) for realizing the benefits of HPC.
• The survey implications are sobering critical
  U.S. supply chains and the leadership of many
  U.S. industries may be at risk if more companies
  do not embrace modeling and simulation with HPC.

C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes
63
Key Study Findings Major Thematic Areas
www.wtec.org/sbes
C. Sagui and S.C. Glotzer
64
Key Findings Life Sciences Medicine

1. Predictive biosimulation is here.
2. Pan-SBES synergy argues for a focused investment
   of SBES as a discipline.
3. Worldwide SBES capabilities in life sciences and
   medicine are threatened by lack of sustained
   investment and loss of human resources.

65
Key Findings Materials

1. Computational MSE is changing how new materials
   are discovered, developed, and applied, from the
   macroscale to the nanoscale.
2. World-class research exists in all areas of
   materials simulation in the US, EU, and Asia the
   US leads in some, but not all, of the most
   strategic of these.
3. The US ability to innovate and develop the most
   advanced materials simulation codes and tools in
   strategic areas is eroding.

66
Key Findings Energy Sustainability

1. In the area of transportation fuels, SBES is
   critical to stretch the supply and find other
   sources.
2. In the discovery and innovation of alternative
   energy sources including biofuels, batteries,
   solar, wind, nuclear SBES is critical for the
   discovery and design of new materials and
   processes.
3. Petascale computing will allow unprecedented
   breakthroughs in sustainability and the
   simulation of ultra-large-scale sustainable
   systems, from ecosystems to power grids to whole
   societies.

67
Key Study Findings Cross-Cutting Issues
www.wtec.org/sbes
C. Sagui and S.C. Glotzer
68
Key Findings Next-generation Architectures and
Algorithms

• Finding 1 The many orders-of-magnitude in
  speedup required to make significant progress in
  many disciplines will come from a combination of
  synergistic advances in hardware, algorithms, and
  software, and thus investment and progress in one
  will not pay off without concomitant investments
  in the other two.

69
Key Findings Next-generation Architectures and
Algorithms

• Finding 2 The US leads both in computer
  architectures (multicores, special-purpose
  processors, interconnects) and applied algorithms
  (e.g., ScaLAPACK, PETSC), but aggressive new
  initiatives around the world may undermine this
  position.
• Already, the EU leads the US in theoretical
  algorithm development, and has for some time.
• Finding 3 The US leads in the development of
  next-generation supercomputers, but Japan,
  Germany committed, and China now investing in
  supercomputing infrastructure.

70
European Initiatives

• A new European initiative called Partnership for
  Advanced Computing in Europe (PRACE) has been
  formed based on the infrastructure roadmap
  outlined in the 2006 report of the European
  Strategy Forum for Research Infrastructures
  (ESFRI 2006). This roadmap involves 15 different
  countries and aims to install five petascale
  systems around Europe beginning in 2009 (Tier-0),
  in addition to national high-performance
  computing (HPC) facilities and regional centers
  (Tiers 1 and 2, respectively). The estimated
  construction cost is 400 million, with running
  costs estimated at about 100200 million per
  year. The overall goal of the PRACE initiative is
  to prepare a European structure to fund and
  operate a permanent Tier-0 infrastructure and to
  promote European presence and competitiveness in
  HPC. Germany and France appear to be the leading
  countries.

71
European Initiatives

• Recently, several organizations and companies,
  including Bull, CEA, the German National High
  Performance Computing Center (HLRS), Intel, and
  Quadrics, announced the creation of the TALOS
  alliance (http//www.talos.org/) to accelerate
  the development in Europe of new-generation HPC
  solutions for large-scale computing systems. In
  addition, in 2004 eleven leading European
  national supercomputing centers formed a
  consortium, DEISA, to operate a continent-wide
  distributed supercomputing network. Similar to
  TeraGrid in the United States, the DEISA grid
  (http//www.deisa.eu) in Europe connects most of
  Europes supercomputing centers with a mix of
  1-gigabit and 10-gigabit lines.

72
Key Findings Scientific Engineering Software
Development

• Finding 1 Around the world, SBES relies on
  leading edge (supercomputer class) software used
  for the most challenging HPC applications,
  mid-range computing used by most scientists and
  engineers, and everything in between.

73
Key Findings Scientific Engineering Software
Development

• Finding 2 Software development leadership in
  many SBES disciplines remains largely in US
  hands, but in an increasing number of areas it
  has passed to foreign rivals, with Europe being
  particularly resurgent in software for mid-range
  computing, and Japan particularly strong on
  high-end supercomputer applications. In some
  cases, this leaves the US without access to
  critical scientific software.

74
Key Findings Scientific Engineering Software
Development

• Finding 3 The greatest threats to US leadership
  in SBES come from the lack of reward,
  recognition and support concomitant with the long
  development times and modest numbers of
  publications that go hand-in-hand with software
  development the steady erosion of support for
  first rate, excellence-based single investigator
  or small-group research in the US and the
  inadequate training of todays computational
  science and engineering students the would-be
  scientific software developers of tomorrow.

75
Key Findings Multiscale Modeling and Simulation

• Finding 2 The lack of code interoperability is
  a major impediment to industrys ability to link
  single-scale codes into a multiscale framework.
• Finding 3 Although U.S. on par with Japan and
  Europe, MMS is diffuse, lacking focus and
  integration, and federal agencies have not
  traditionally supported the development of codes
  that can be distributed, supported, and
  successfully used by others.
• Contrast with Japan and Europe, where large,
  interdisciplinary teams are supported long term
  to distribute codes either in open-source or
  commercial form.

76
Key Findings Engineering Simulation

• Finding 1 Software and data interoperability,
  visualization, and algorithms that outlast
  hardware obstruct more effective use of
  engineering simulation.
• Finding 2 Links between physical and system
  level simulations remain weak. There is little
  evidence of atom-to-enterprise models that are
  coupled tightly with process and device models
  and thus an absence of multi-scale SBES to
  inform strategic decision-making directions.

77
Key Findings Engineering Simulation

• Finding 3 Although US academia and industry are,
  on the whole, ahead (marginally) of their
  European and Asian counterparts in the use of
  engineering simulation, pockets of excellence
  exist in Europe and Asia that are more advanced
  than US groups, and Europe is leading in training
  the next generation of engineering simulation
  experts.

78
Key Findings Validation, Verification
Uncertainty Quantification

• Finding 1 Overall, the United States leads the
  research efforts today, at least in terms of
  volume, in quantifying uncertainty however,
  there are similar recent initiatives in Europe.

79
Key Findings Validation, Verification
Uncertainty Quantification

• Finding 2 Although the U.S. DOD and DOE are been
  leaders in VV and UQ efforts, they have been
  limited primarily to high-level systems
  engineering and computational physics
  mechanics, with most of the mathematical
  developments occurring in universities by small
  numbers of researchers. In contrast, several
  large European initiatives stress UQ-related
  activities.
• Finding 3 Existing graduate level curricula,
  worldwide, do not teach stochastic modeling and
  simulation in any systematic way.

80
Key Findings Big Data, Visualization, and
Data-Driven Simulation

• Finding 1 The biological sciences and the
  particle physics communities are pushing the
  envelope in large-scale data management and
  visualization methods. In contrast, the chemical
  and material science communities lag in
  prioritization of investments in data
  infrastructure.
• Bio appreciates importance of integrated,
  community-wide infrastructure for massive amounts
  of data, data provenance, heterogeneous data,
  analysis of data and network inference from data.
  Great opportunities for the chemical and
  materials communities to move in a similar
  direction, with the promise of huge impacts on
  the manufacturing sector.

81
Key Findings Big Data, Visualization, and
Data-Driven Simulation

• Finding 2 Industry is significantly ahead of
  academia with respect to data management
  infrastructure, supply chain, and workflow.

82
Key Findings Big Data, Visualization, and
Data-Driven Simulation

• Most universities lack campus-wide strategy for
  big data.
• Widening gap between the data infrastructure
  needs of the current generation of students and
  the campus IT infrastructure.
• Industry active in consortia to promote open
  standards for data exchange a recognition that
  SBES is not a series of point solutions but
  integrated set of tools that form a workflow
  engine.
• Companies in highly regulated industries, e.g.,
  biotechnology and pharmaceutical companies, are
  also exploring open standards and data exchange
  to expedite the regulatory review processes for
  new products.

83
Key Findings Big Data, Visualization, and
Data-Driven Simulation

• Finding 3 Big data and visualization
  capabilities are inextricably linked, and the
  coming data tsunami made possible by petascale
  computing will require more extreme visualization
  capabilities than are currently available, as
  well as appropriately trained students who are
  adept with data infrastructure issues.
• Finding 4 Big data, visualization and dynamic
  data-driven simulations are crucial technology
  elements in grand challenges, including
  production of transportation fuels from the last
  remaining giant oil fields.

84
Inadequate education training threatens global
advances in SBES a worldwide concern

• Insufficient exposure to computational science
  engineering and underlying core subjects at high
  school and undergraduate level
• Increased topical specialization beginning with
  graduate school
• Insufficient training in HPC an educational
  gap
• Gap b/t domain science courses and CS courses
  insufficient continued learning opportunities
  related to programming for performance
• Major worry for multicore/gpu architectures in US
• Students use codes as black boxes who will be
  innovators?
• No real training in software engineering for
  sustainable codes
• Little training in UQ, VV, risk assessment
  decision making
• Necessary for atoms to enterprise US lead slim

S.C. Glotzer 01/29/09
MASI Conference, Helsinki, Finland
www.wtec.org/sbes
85
Key Findings Education Training

• Finding Continued progress and U.S. leadership
  in SBES and the disciplines it supports are at
  great risk due to a profound and growing scarcity
  of appropriately trained students with the
  knowledge and skills needed to be the next
  generation of SBES innovators.
• Current background training is insufficient.
• The U.S. lead in many areas is decreasing across
  all SE indicators
• U.S. doctorates in SE lt EU or Asia.
• Fierce competition for international recruiting.
• New interdisciplinary education programs in EU

86
Opportunities for the US to gain or reinforce
lead in SBES
www.wtec.org/sbes
C. Sagui and S.C. Glotzer
87
Opportunities for the US to gain or reinforce
lead in SBES

• Finding 1 There are clear and urgent
  opportunities for industry-driven partnerships
  with universities and national laboratories to
  hardwire scientific discovery to engineering
  innovation through SBES.
• This would lead to new and better products, as
  well as development savings both financially and
  in terms of time.
• National Academies report on Integrated
  Computational Materials Engineering (ICME), which
  found a reduction in development time from 10-20
  yrs to 2-3 yrs with a concomitant return on
  investment of 31 to 91.

88
Opportunities for the US to gain or reinforce
lead in SBES

• Finding 2 There is a clear and urgent
  opportunity for new mechanisms for supporting
  SBES RD.
• Support and reward for long-term development of
  algorithms, middleware, software, code
  maintenance and interoperability.
• Although scientific advances achieved through the
  use of a large complex code is highly lauded,
  the development of the code itself often goes
  unrewarded.
• Community code development projects are much
  stronger within the EU than the US, with national
  strategies and long-term support.
• investment in math, software, middleware
  development always lags behind investment in
  hardware

89
Opportunities for the US to gain or reinforce
lead in SBES

• Finding 3 There is a clear and urgent
  opportunity for a new, modern approach to
  educating and training the next generation of
  researchers in high performance computing for
  scientific discovery and engineering innovation.
• Must teach fundamentals, tools, programming for
  performance, verification and validation,
  uncertainty quantification, risk analysis and
  decision making, and programming the next
  generation of massively multicore architectures.
  Also, students must gain deep knowledge of their
  core discipline.

90
For more information and final reportwww.wtec.or
g/sbes
C. Sagui and S.C. Glotzer
SIAM, CSE09
www.wtec.org/sbes