Brent Fultz; Co-PIs are Michael Aivazis, Ian Anderson; PM is Mike McKerns

1
Brent Fultz Co-PIs are Michael Aivazis, Ian
Anderson PM is Mike McKerns California Institute
of Technology
2
Spallation Neutron Source (SNS)
Accelerator Target Instruments B 1.411
3
Reasons for DANSE
New Science Better Science Ease of Use Software
Stability and Reuse Support Early Operations of
the SNS
4
Excitations in Solids An Optical Illusion
of Space-Time Correlations
5
Neutron Scattering
Word Pairs Examples
coherent, elastic coherent, inelastic incoher
ent, elastic incoherent, inelastic
diffraction dispersions diffuse
scattering densities of states
6
(No Transcript)
7
elastic
8
inelastic
9
Coherence Preserves the Phase
10
Coherence
11
Q

k
k
12
Incoherence Arbitrary Phase
13
Incoherence
14
Structure and Dynamics
http//physics.ucsc.edu/research/images/
15
Structure in Space Dynamics in Time
16
Diffraction
17
Inelastic (incoherent, inelastic) vs. (coherent,
inelastic)
18
Correlation Functions
Diffraction, SANS gives spatial correlation
functions (Patterson function
structure) Inelastic scattering gives space-time
correlation function (Van Hove correlation
function dynamics)
19
reduction and visualization
20
(No Transcript)
21
The DANSE Project
22
DANSE Project Big Picture

• Enable new neutron scattering science by
  scientific computing.
• Allow experimentalists to rearrange modularized
  components, and create new components.
• Build enough important and quality pieces to make
  DANSE the choice of future developers.
• Well-defined tasks organized by subfields.
• System in place for Earned Value Management.
• M 12 over 5 years

23
Operational Goals

• Provide todays capabilities of reduction and
  visualization
• Enable new types of science in major subfields
• Standardize software design, and start to
  standardize usage
• Provide a runtime framework onto which scientists
  can add components
• Support for early SNS instruments
• Maintainable by SNS before the end of the project

24

25
Component Framework
Encapsulate scientists code within component. A
component inherits methods from the
framework, user communicates properties through
framework.
26
Users Differ in Needs and Expertise
1. Beginning Student 2. Senior Student or
Postdoc 3. Young Scientist 4. Established
Researcher 5. Instrument Scientist 6. Software
Developer 7. System Maintainer
One software system must serve everyone, but
different people need different interfaces.
27
Users 1. Beginning Student

• Complexity reduced to essential modules
• Reasonable defaults for most choices
• Failures must be diagnosed and explained
• Help with science principles

Dashboard-type GUI and Help services
28
Users 3. Young Scientist

• Flexible, interactive explorations of data
• Modeling and simulation packages
• Compare outputs of different analyses
• Software must facilitate science

Seeks empowerment from GUI and command-line
control.
29
Users 6. Software Developer

• Well-documented access to the integration layer
• Portable building tools
• Robust framework for debugging
• Regression test suites
• Development process and tools for quality and
  efficiency

Direct command line control, but support for all
other user interfaces.
30
(No Transcript)
31
When analysis components are building blocks,
what will scientists build?
32
Facilitate discovery in neutron scattering
science by raising computation to a higher level
of abstraction.
Can scientific careers be developed by adding
value at this higher level?
33
(No Transcript)
34
Deployment Diagrams

35
theorist_at_mit
colleague_at_caltech
data_at_ornl
Experiment teams are already delocalized
geographically. Natural for collaborative
interactions across distributed computing
resources
system_at_sns
36
DANSE and the NSF

• Leverage the DOE BES investment in SNS hardware
  to do innovative science (often done under NSF
  grants)
• Software has strong connections to NSF emphasis
  on
• scientific principles
• education
• Initiate a partnership between a university-based
  developer community and a national facility
• Communication and science community outreach is
  central to DANSE should promote NSF-DOE
  interactions too
• DANSE, both project and product, is extensible to
  other facilities and disciplines

37
DANSE and the SNS

• DANSE project augments SNS expertise and
  resources
• Early DANSE deliverables track instrument
  commissioning at the SNS
• DANSE project will facilitate new science from
  the SNS, especially in the early years of
  operations
• Good communication channels between SNS and DANSE
• Decoupling of software development efforts
• - DANSE emphasis on advanced data analysis
  (modeling and simulation)
• - SNS emphasis on data services (acquisition and
  curation)
• Engineering consistency SQRL consults for the
  SNS, subcontract with DANSE

38
DANSE Project
39
DANSE Effort Today (Personnel)

• All senior personnel together since early 2003
• scientific and technical leadership
• assessment of what is practical to deliver
• Professional staff
• infrastructure, support tools, institutional
  memory
• Postdoc Software Developers
• mostly physical scientists interested in software
• varying software skill okay, but cannot start at
  zero
• must sustain their scientific research
• transitions out of DANSE/ARCS projects have been
  successful
• Graduate Students
• can leverage DANSE in thesis research
• testing and some development
• Undergraduate Students
• mostly in computer science
• External collaborators

40
DANSE Effort Today (Technical)

• Scientific capabilities defined, packages
  selected
• (over 4 years, 12 workshops, 6 polls,
  and former CED project)
• ARCS inelastic software in production, pyre
  framework
• Reflectometry software in production, moving to
  framework
• Software development process is established
• Common components for I/O, interface, and
  numerics
• gives economies of scale, reusability,
  uniformity,
• and fewer bugs

41
DANSE Effort Today (Management )

• Earned value management established with a
  Project Baseline
• Change, risk, configuration control plans are
  ready
• Infrastructure tools identified (svn, trac,
  build/release, gui)
• Development processes are specified
• Project management in place

42
Work Breakdown Structure

• Think of Microsoft Project
• Layout of all tasks. Each task has
• - description in a dictionary file
• - estimate of time
• - estimate of cost
• - markers of task progress
• WBS maintained by Mike McKerns, Project Manager
• Changes are expected, but must be documented.

43
Earned Value Management System

• Beyond Microsoft Project, becoming a federal
  requirement
• Monthly reports of progress (Earned Value)
• - use markers of task progress
• Monthly reports of expenditures (Actual Costs)
• - estimates of time and cost
• Actual Costs and Earned Value are in units of
  dollars
• EVMS maintained by B. Thibadeau at SNS

44
Integration ofDesign Process, QA, Releases

• State purpose of component, specifications,
  functional tests
• Object-oriented programming allows early
  selection of design patterns
• Review, commit to svn
• Prototype. Review the design and costs
• Build to specifications, revise specs, develop
  certification tests
• Deploy to all target systems

45
Features of the Development Process

• Misperceptions of specifications are the biggest
  source of risk in software projects.
• DANSE developers write specs, so communication
  problems are minimized.
• Code inspections and QA practices further
  minimize risk.
• Agile commercial practices are appropriate for a
  scientific development team.
• Good specifications and requirements can
  facilitate collaborations.

46
Education

• Pre-service teacher education at ISU
• develop K-12 lesson plans from subprojects
• adoption by young teachers
• Nanoscience program at MSU
• Student involvement
• minority student funds
• testers of software and documentation
• some in computer science
• New role for documentation (texts by Fultz,
  Billinge, Ustundag)

47
(No Transcript)
48
Broader Impact

• Algorithm Development Moves to Higher Levels
• flexible integration of DANSE components and
  applications
• subproject science packages
• DANSE should be Useful for Other Groups
• other neutron facilities
• synchrotron radiation research
• education programs in materials/chemistry/condense
  d matter
• Demonstrate a New Approach to Scientific Software
  Development
• professional standards with O-O design
• distributed team, now cohesive
• some interest from computational scientists

49
Summary of DANSE

• The road from neutron data to neutron science is
  paved by computing.
• DANSE system moves data analysis at the a higher
  level of abstraction where new science is found.
• Integration of Design, Development, QA, Release
  Management.
• Plan, key personnel, technical and management
  systems are in place for construction of DANSE.
  Now funded by NSF IMR-MIP program 2006-2011.

50
For this Kickoff Meeting

• DANSE Subproject Teams
• - Status reports
• - Technical discussion
• Collaborators
• - Look over DANSE status and plans
• - We want to hear updates from you too
• Visitors to this Kickoff Meeting
• - Do we have common interests?
• - Keep in touch!

51

• End Presentation