Grids and the Virtual Observatory

1
Grids and the Virtual Observatory

• Original topic by Roy Williams
• Presentation by Allen Harvey

2
The Virtual Observatory

• A massive, heterogeneous repository of datasets
  to be analyzed/studied
• Distributed over a computational grid
• Pieces all around the world (primarily the US)
• US, UK, Europe
• Designed to be a Rosetta Stone
• Convert astronomers from specialists in a tool or
  wavelength to a specialist in a process or
  phenomenon
• Federate Data Bring all data into the same
  frame of reference, i.e. change apples and
  oranges into pears.

3
Grids

• Emphasis on data collection and storage
• Services tend to be internet based
• Use of GSI Authentication, Storage Resource
  Broker, GridFTP
• Grids will storage many terabytes of data
• Store finished calculations to improve performance

4
Databases

• Collection of Images
• Gamma, X-Ray, Optical, infrared, etc.
• Collection of measurements
• Positions, brightness, angles, etc.
• Collection of studies
• Many terabytes
• Sloan Digital Sky Survey 2-15 TB
• Heavily used for statistical astronomy

5
Database Challenges

• Performing Queries
• Sharing of databases
• VOTable Standard
• High performance binary
• Metadata can be too large
• Minimize XML usage
• Need very large caches Queries need data about
  the records they are performed on
• Security accessing different databases are
  parts of databases in different locations

6
Virtual Sky

• Created by Caltech Univ., Johns Hopkins Univ.,
  Sloan Sky Survey, and Microsoft Research.
• Uses images. Some sources include
• Sloan Digital Sky Survey
• Digital Palomar Observatory Sky Survey
• Hubble Deep Field
• The NOAO Deep Wide Field Survey
• Images are resampled to a standard projection
  (Google Earth works this way)
• Accessed from http//virtualsky.org
• Many images have been preprocessed already
• Output files are spread uniformly over processors
• As of 2003, the code was being ported to the
  TeraGrid

7
SkyView
8
Montage

• Mosiacking software by NASA and Caltech
  University
• Simultaneous parallel processing of images of
  different wavelengths
• Pulls data from most convenient location
• Uses virtual data data as-is or data that needs
  to be computed first
• Allows for custom input
• User can define fitting functions, filters,
  estimation levels, etc.
• Saves work others have done so future work is
  faster
• Think of this as a more powerful Virtual Sky

9
Possible Studies

• Stacking rule out noise in images by comparing
  the same space from different sources
• Spectrophotometry merging detections from
  various wavelengths
• Extended Sources Study objects that vary in
  shape with wavelength
• Image differencing subtracting bands of
  wavelengths from other images
• Time Federation see how an object changes in
  time
• Essentially Multiwavelength Objects discover
  objects only visible at certain wavelengths

10
Teragrid
11
Semantics

• Need for standard ontologies
• Books
• QB6 Star Catalogs
• QB349-QB480 Theoretical Astronomy and
  Astrophysics
• Virtual Observatory
• Strasbourg Ontology Unified Content Descriptor
  based on a hierarchical classification system and
  was the first one created
• ???

12
Semantics (Cont.)

• Not universally agreed upon
• Need to evolve in time
• Must be able to map from one ontology to another
• Research is ongoing
• Try to treat ontologies as namespaces
• Data may be binary or text
• Must allow for modification of metadata
• Ex. Type A Type B Type A-B

13
Conclusions

• Grid applications are evolving
• Can benefit from greater collaboration Not just
  NASA and a few universities
• Much of the data and technology are ready, just
  needs implementation
• Data access is the bottleneck