2001 Euresco Conference on PSEs: Past, Present and Future

1
2001 Euresco Conference on PSEsPast, Present
and Future
Tony Hey Director UK e-Science Core
Programme EPSRC, Swindon UK Tony.Hey_at_epsrc.ac.uk
2
Outline

• Review of Presentations
• Summary of Key Trends
• Focus and Format of next Euresco PSE Conference

3
Review of Presentations (1)

• David Walker
• - Vision for PSEs - Grids
• - Grid Computing Environments
• - Web Portals
• - New technologies XML, XSLT
• - Jini, Agents and the Semantic Web
• - Application Service Providers

4
Review of Presentations (2)

• Geoffrey Fox
• - Grid vs P2P architecture
• - Shawn Fanning of Napster
• (75M users cant be wrong)
• - .antiNet
• - JXTA as Unix Shell for Web OS
• - Grid as Enterprise P2P
• - Grid Portals for Education

5
Review of Presentations (3)

• Kai-Uwe Schmitt
• - State of the Art of Car Crash
• Simulation Software
• - Industrial distrust of Internet
• - PSEs not really on the agenda

6
Review of Presentations (4)

• Elias Houstis
• - GasTurbnLab PSE
• 31,000 parts
• 10,000 parameters
• 100,000 s/w modules
• - 3 tier architecture with Grasshopper
• Agent-based middleware
• - Methodology requires access to source
• code

7
Review of Presentations (5)

• Young Researcher Presentations
• - Most not about PSEs but about
• computational science software
• - Lack of comparison/awareness
• with related research
• - Discussion of Jini and CORBA
• technology

8
Review of Presentations (6)

• Hans Petter Langtangen
• - Re-engineering multi-physics codes
• - OO Design, Rapid Prototyping
• - Diffpack approach and maintainability
• - Success with Daimler-Chrysler codes
• - Performance of Parallel versions?

9
Review of Presentations (7)

• Chris Johnson
• - Computational Science Pipeline
• Modelling Simulation Visualization
• - Need ability to interact anywhere
• - SCIRun, BIOPSE examples
• - DOE C-SAFE Center
• Hard problems, scale to 2000 PEs
• - NIH Center
• Real success with immersive
• visualization technology

10
Review of Presentations (8)

• Jeremy Wu
• - CFD Colourful Fluid Dynamics!
• - Pay-as-you-go Software Licences
• - CFD Application Service Provider
• - Impact of the Grid?

11
Review of Presentations (9)

• Peter Sloot
• - Polder Grid Infrastructure
• - Virtual Laboratory Environment
• - Lattice Boltzmann Blood Flow Example
• - Measurements in VR?
• - Need for Domain Knowledge
• Computer Science
• - Component-based approach to PSEs

12
Review of Presentations (10)

• Hans-Christian Hoppe
• - Users in Stone Age!
• - UNICORE Project
• Good interface
• Incarnation Database
• - EUROGRID Project
• BIOGRID, METEOGRID,
• CAE-GRID, HPC-GRID
• - VHS versus Betamax analogy
• Move to Globus/GGF standard ASAP!

13
Review of Presentations (11)

• Omer Rana
• - Human Usability Issues - Pancaked!
• - Wrapping technologies for heritage codes
• - Impact of the Grid
• Think service not code
• - Agent service approach
• - New security issues?

14
Review of Presentations (12)

• Klaus Wolf
• - Autobench and TENT
• - Realistic Industrial Workflow
• CAD - multiple CAE packages
• - Data Integration key issue
• Wrapping non-trivial problem
• - Use OO Data Structures and CORBA
• - Need for CAD Repair Man!
• - Hard to build industrial-strength PSE

15
Review of Presentations (13)

• Yike Guo
• - PSEs for Knowledge Discovery
• - High Throughput or Data Intensive?
• - Data Information - Knowledge
• - Generic Grid-like Architecture
• - IP logging as e-Book
• - Applications in Biology, Chemistry,

16
Preview of Presentation

• Jason Jones
• - PSUE Parallel Simulation User
• Environment projects with BAE Systems
• - EU CAESAR and JULIUS projects
• - Computational Science Pipeline
• Mesh CFD Refinement - Visualization
• - Data transfer between programmes is key
  problem
• - Use CORBA and MEM-COM D/B?

17
Size and Complexity of Industrial Applications

• Systems with thousands of mechanical
  components
• e.g. 31000 in Gas Turbine
• Large numbers of collaborating people
• e.g. LHC experiments with 1000 physicists
• Large grid sizes for aircraft simulation
• - Viscous and Turbulent flow 5-8 million
• - Electromagnetic scattering 12-16 million

18
(No Transcript)
19
(No Transcript)
20
TENT Architecture
21
Some Comments on CORBA

• Many systems CORBA based but specific ORB rarely
  mentioned
• Advantage of CORBA One specification, 800
  companies in the consortium
• Disadvantage Large number of implementations
  and vendors offering various ORBs (Orbacus,
  Orbix, )
• Using different ORBs in the same system is not
  recommended!
• Programming using CORBA is not trivial, the C
  syntax is rather complex

22
Corba Middleware OBACUS

• Stable commercial product
• Available source code for modifications if
  required
• Support for IDL Java and C mapping
• Services
• Event server
• Name server
• Trading server
• Implementation Repository
• Interface repository

23
Some comments on PSEs

• Varying architectures, inter-process
  communications, non-standard interfaces, poor
  modularity
• Data flow between CAD and CAE Tools major
  problem
• No generally accepted methodology for
  specification, analysis and design of modular
  reusable systems
• Lack of standards to support commonality and
  modularity of systems/components
• Progress should be measured in the number and
  quality of reusable, robust design tools.
• Problem of vendor specific legacy tools and
  codes

24
Future PSEs will offer support for experimental
work?

• Present PSEs focus on design side of the
  process -
• experimental side not supported.
• Need to incorporate validation and testing
  stages into PSE
• Support for recording and analysis of
  experimental data
• Incorporate databases of experimental
  measurements into PSE.
• Will allow Data-mining of experimental and
  simulation data
• Convergence with knowledge discovery PSEs?

25
Some Comments on Technologies

• XML, SOAP and UDDI cannot be ignored
• - need to link to Web Services world
• - XML schemas as Objects
• Standard 3 Tier Architecture challenged by
• P2P and Users regarded as Resources
• - JXTA?
• Agents will come of age
• - Ontologies
• - DAML/OIL
• - Semantic Web
• The Grid is Coming!

26
Grid CORBA extensions?
27
Format and Focus of Next Conference?

• Chris Johnsons comments on Format
• Fewer talks!
• - A couple of overview talks
• - A few state of the art PSE talks
• - Longer or more frequent breaks
• Some Panel discussions with lots of audience
  participation

28
Format and Focus (2)

• Suggested Panel Topics
• - Outstanding problems and issues
• - PSE Developers vs PSE Users
• - Controversial topic such as
• Are PSEs worthwhile?
• - Software Architecture design and evaluation
• - How move forward science or engineering of
  PSE development?

29
Suggested Focus e-Science

• Many e-Science/Grid Testbeds will have results to
  report
• Grid Technologies should be maturing
• Globus, SRB, JXTA - ?
• Should be examples of new Application Service
  Providers based on Grid/Web technologies

30
UK e-Science Testbeds

• Particle Physics Grid linked with EU DataGrid and
  US GriPhyN projects
• Virtual Observatory AstroGrid
• Engineering Grids
• Chemistry/Materials Grids
• Bioinformatics Grid
• Medical and Healthcare Informatics Grids
• Environmental Grids

31
A Grid IRC Grand Challenge

• DIRC -Verification and validation of safety
  critical and high integrity systems
• AKT - Integrated set of technologies, methods
  and services to support knowledge life cycle
• EQUATOR - Development of theories and devices
  integrating physical and digital worlds
• Medical Images to Clinical Information
• Medical and Healthcare Informatics as a Grand
  Challenge Application for the Grid

32
Other e-Science Grid Projects

• US Projects
• GriPhyn, NEESGrid, NVO, ..
• EU Projects
• DataGrid, DataTAG (TTT)
• MeteoGrid, HPCGrid, CAEGrid
• Other National Projects
• PolderGrid,