Alastair Hume (EPCC, University of Edinburgh), Amy Krause (EPCC, University of Edinburgh)
Mark Holliman (Institute for Astronomy, University of Edinburgh)
Robert G. Mann (Institute for Astronomy, University of Edinburgh)
Keith Noddle (Institute for Astronomy, University of Edinburgh)
Stelios Voutsinas (Institute for Astronomy, University of Edinburgh)
Abstract
Querying distributed datasets has been a long held but elusive goal of the International Virtual Observatory Alliance (IVOA). Difficulties include designing a service infrastructure that can handle the necessary tasks of metadata federation, data access and query planning across an arbitrary set of independent data sources. We have now built a service that accomplishes this utilizing the IVOA Table Access Protocol (TAP) in conjunction with the database oriented grid middleware OGSA-DAI. It utilizes a three layer architecture to provide data access to an arbitrary collection of TAP services.
Users submit a list of TAP endpoints along with their associated VOSI table metadata locations to our TAP Service Federation Factory. The Factory returns a TAP endpoint for the newly generated service that federates the requested datasets. Users can then query this new TAP service as they would any other, be it through command line software or GUI clients like Topcat.
To generate this TAP service, the Federation Factory retrieves and combines the metadata into an OGSA-DAI resource. It then deploys a TAP service on top of the OGSA-DAI resource connected through JDBC. The submitted queries can contain JOIN operations across any of the federated tables; OGSA-DAI handles the complex tasks of query planning and processing to ensure the queries run efficiently.
In developing this service we have identified several TAP metadata additions that would facilitate better query planning and execution. These will be presented to the IVOA DAL Working Group for inclusion in a future version of the TAP protocol.
We are currently exploring the best approach for deploying our TAP Service Federation Factory in a production role for use by the general astronomical community. A promising start has been made as part of an MSc project utilising a Web2.0 interface to a Python framework that makes querying distributed datasets both easy and intuitive.
An online demo of TAP Service Federation Factory can be found at: http://www.ogsadai.org.uk/demos/astro/
Poster in PDF format
Paper ID: P059
Poster Instructions
|