Report from DANSE Workshop

1
Report from DANSE Workshop

• Sept. 3-8, 2003
• Goals
• 1) To explain DANSE to selected scientists and
  engineers who develop software for neutron
  scattering.
• 2) To start organizing a national project.

2
Attendees

• Persons
• Michael Aivazis, Simon Billinge, Paul Butler,
  Brent Fultz, Garrett Granroth, Gayle Greene, Nick
  Hauser, Jason Hodges, Tim Kelley, Paul Kienzle,
  June Kim, Przemek Klosowsky, JaeDong Lee, Mark
  Lumsden, Mike McKerns, Ray Osborn, Peter
  Peterson, Lee Robertson, Frans Trouw, Ersan
  Ustundag, Xun-Li Wang, Mohana Yethiraj
• Specialties
• diffractometry, SANS, inelastic scattering,
  engineering diffraction, reflectometry, software
  engineering

3
Reasons to do DANSE

• 1. Better Science
• - More science (efficiency, use of beamtime)
• - New science (extensibility)
• - Connection to modern theory
• 2. Ease of Use
• Cases from casual to expert users
• Leveraging knowledge of senior scientists
• 3. Ease of Development
• Cases from casual to expert users
• 4. Cost Effective for Neutron Facilities
• - less duplication of effort, easier user
  support, universal look and feel
• 5. Time is Right
• Grid is coming and needs scientific applications
• Potential for growth internationally, and into
  other fields such as x-ray techniques
• Framework from ASCI
• SNS is a fresh opportunity to do coherent
  software

4
What Is the DANSE Architecture?

• An integration system to glue together data
  analysis software, offering
• - Ease of use at all levels of user expertise
• - Extensibility (modular)
• Distributed and standalone computing
• An open source project

5
What Is the DANSE Project?

• The incorporation of present software
  capabilities into a broader system, and sometimes
  using the existing packages themselves (DAVE,
  ISAW)
• The development of powerful software to enable
  new science from the instruments at the SNS
• A collaborative effort between software
  professionals, neutron scattering scientists, and
  facilities
• An open source project

6
What DANSE is Not

• A barrier to using existing tools
• An exercise in high-performance computing
• A rewrite of DAVE, ISAW, or other packages
• Dependent on a remote server
• More money for the ARCS instrument

7
DANSE Project Has Two Parts

• 1. Science-Based Subprojects
• Inelastic (ARCS)
• Diffraction
• Engineering Solid Mechanics
• Reflectometry
• SANS
• 2. An Infrastructure of Central Services
• Software Framework
• Integration of Science Sub-Projects
• Hardware

8
Organizational Structure of DANSE Project
9
Science Sub-Projects Shall

• Prioritize Software Needs
• Ensure Availability of Existing Capabilities
• Rebuild Them as Needed for Ease of Use
• Envision and Build New Approaches for Data
  Analysis
• Strong Preference for Real Data from an Existing
  Instrument

10
Central Services Shall

• Coordinate and Integrate the Products of the
  Scientific Subfields
• Design and build
• - Modular Components
• - Data Streams
• - Distributed Analysis System and compatibility
  with standalone capabilities
• Prototype Hardware and Software Facilities
• Educational and Consulting Role for Subprojects
• Release Management

11
Programming Languages

• The following can be used within DANSE
• Python
• C, C
• Fortran
• Java, Tcl
• Environments (Matlab, IDL, Igor)
• New codes will be developed in
• Python, C, C, Fortran

12
User/Developer Interfaces

• GUI 1 (basic buttons)
• GUI 2 (Visual Programming Environment)
• Python command line (analysis packages)
• Developer tools

13
DANSE Steps for Sub-Project Leaders

• Prioritize the existing analysis methods in your
  field
• Informed consensus is best, but guesses may
  suffice
• Emphasize analysis needs, not means to the ends
• Develop a budget for a 3-person team
• Write a 5-page sub-proposal
• Expect iteration.

14
DANSE Steps for Brent and Michael

• Rewrite the DANSE white paper (more Why, less
  What)
• Develop a full project plan (Scope. Effort
  estimation, e.g. graphics)
• Management plan (scenarios depend on funding)
• Early distributions of
• - stream protocols for histograms, tables, and
  XML data
• - plan for server installation of a Python
  package from each sub-group
• - DANSE client software distribution plan
• Coordinate with Ian the letters of support and
  visits to NSF and DOE