HR: 14:55h
AN: IN53C-06 [Abstracts]
TI: Metadata Management on the SCEC PetaSHA Project: Helping Users Describe, Discover, Understand, and Use Simulation Data in a Large-Scale Scientific Collaboration
AU: * Okaya, D
EM: okaya@usc.edu
AF: Dept Earth Sciences, Univ. Southern California, Los Angeles, CA 90089-0740, United
States
AU: Deelman, E
EM: deelman@isi.edu
AF: USC Information Sciences Institute, 4676 Admiralty Way, Marina del Rey, CA 90292, United
States
AU: Maechling, P
EM: maechlin@usc.edu
AF: SCEC, Univ. Southern California, Los Angeles, CA 90089-0740, United States
AU: Wong-Barnum, M
EM: mona@sdsc.edu
AF: San Diego Supercomputer Center, UC San Diego, San Diego, CA 92093-0505, United
States
AU: Jordan, T H
EM: tjordan@usc.edu
AF: SCEC, Univ. Southern California, Los Angeles, CA 90089-0740, United States
AU: Meyers, D
EM: dpmeyers@usc.edu
AF: SCEC, Univ. Southern California, Los Angeles, CA 90089-0740, United States
AB:
Large scientific collaborations, such as the SCEC Petascale Cyberfacility for Physics-based Seismic Hazard
Analysis (PetaSHA) Project, involve interactions between many scientists who exchange ideas and research
results. These groups must organize, manage, and make accessible their community materials of observational
data, derivative (research) results, computational products, and community software. The integration of scientific
workflows as a paradigm to solve complex computations provides advantages of efficiency, reliability,
repeatability, choices, and ease of use. The underlying resource needed for a scientific workflow to function and
create discoverable and exchangeable products is the construction, tracking, and preservation of metadata.
In the scientific workflow environment there is a two-tier structure of metadata. Workflow-level metadata and
provenance describe operational steps, identity of resources, execution status, and product locations and names.
Domain-level metadata essentially define the scientific meaning of data, codes and products. To a large degree
the metadata at these two levels are separate. However, between these two levels is a subset of metadata
produced at one level but is needed by the other. This crossover metadata suggests that some commonality in
metadata handling is needed.
SCEC researchers are collaborating with computer scientists at SDSC, the USC Information Sciences Institute,
and Carnegie Mellon Univ. in order to perform earthquake science using high-performance computational
resources. A primary objective of the "PetaSHA" collaboration is to perform physics-based estimations of strong
ground motion associated with real and hypothetical earthquakes located within Southern California.
Construction of 3D earth models, earthquake representations, and numerical simulation of seismic waves are
key components of these estimations. Scientific workflows are used to orchestrate the sequences of scientific
tasks and to access distributed computational facilities such as the NSF TeraGrid. Different types of metadata
are produced and captured within the scientific workflows. One workflow within PetaSHA ("Earthworks") performs
a linear sequence of tasks with workflow and seismological metadata preserved. Downstream scientific codes
ingest these metadata produced by upstream codes. The seismological metadata uses attribute-value pairing in
plain text; an identified need is to use more advanced handling methods. Another workflow system within
PetaSHA ("Cybershake") involves several complex workflows in order to perform statistical analysis of ground
shaking due to thousands of hypothetical but plausible earthquakes. Metadata management has been
challenging due to its construction around a number of legacy scientific codes. We describe difficulties arising in
the scientific workflow due to the lack of this metadata and suggest corrective steps, which in some cases
include the cultural shift of domain science programmers coding for metadata.
DE: 0525 Data management
DE: 1706 Computational geophysics
DE: 7212 Earthquake ground motions and engineering seismology
DE: 7219 Seismic monitoring and test-ban treaty verification
SC: Earth and Space Science Informatics [IN]
MN: 2007 Fall Meeting