Supporting Large-Scale Science with Workflows

1
Supporting Large-Scale Science with Workflows

• Deana Pennington
• University of New Mexico
• Long-Term Ecological Research Network Office
• ITR Science Environment for Ecological Knowledge
  (SEEK) project
• CI-Team Advancing CI-Based Science through
  Education, Training, and Mentoring of Science
  Communities
• WORKS 07
• June 25, 2007

2
Scientific Research Cycle
Theory
Hypothesis
Research Design
Experiment
Results
Inference
3
System of interest
Causes
Consequences
Climate Population Change
Vegetation Composition Structure
4
Data Flow heterogeneous datasets/models/workflow
s
Plant Dispersal
Species Invasion
Carbon
Plant Growth
Biota
Climate Change Species Distribution
Wildfire
5
Metaprovenance

• Provenance dataset derivation explicit
  information about which workflow components were
  used
• Metaprovenance dataset derivation capture
  tacit information about why those components were
  used and which components go together

6
Many output datasetsComplex workflows/parameter
sweeps
For I 1 to N Climate scenarios
Plant Growth Workflow
For k 1 to N Parameter sets
Climate Workflow
System Workflow
Wildfire Workflow
Other Subsystem Workflows
For j 1 to N Algorithms
7
Metaprovenance

• Project coordination
• Workflow gt 1000 datasets
• Parameter sweep gt 100 parameter sets
• Which dataset do I go to to see???
• Provenance Given a dataset, what
  components/parameters were used?
• Metaprovenance Given a set of
  components/parameters, which dataset was produced?

8
Science Dashboard?

• Enter project level information project
  approach and design
• Control parameters

9
Design Flow
Executable Workflow
10
Knowledge Flow
Formal Ontology
Abstract Workflow
Ontology-driven Workflows
Executable Workflow
11
Knowledge-Driven Workflows
Data Analysis
Inference
Theory
Empirical Results
Experimental Design
Hypothesis testing
12
Acknowledgments

• This work was heavily influenced by discussion
  within the SEEK project and especially the SEEK
  Knowledge Representation team. I appreciate all
  of their interaction. Only my own perspective is
  expressed, and they would not necessarily agree.
  The work was supported by National Science
  Foundation grant 0225665 for the Science
  Environment for Ecological Knowledge (SEEK)
  project and grant 0636317 for the CI-Team
  Demonstration Project Advancing
  Cyberinfrastructure-Based Science through
  Education, Training, and Mentoring of Science
  Communities.