Wednesday
4 February 2004 at 16:15
over the Access Grid nodes located:
at the
University of Leeds:
in the Informatics Research Institute, Conference Room 6.08, Level 6
of the E.C. Stoner Buildingand
at the
University of Sheffield:
in room K17, the Hicks Building (entrance at the top of Hounsfield
Road, Sheffield 10)
A B S T R A C T
The
AstroGrid's Grid Data Warehouse
The
Virtual Observatory movement (www.ivoa.net),
of which AstroGrid (www.astrogrid.org)
is the British contribution, aims to allow processing of distributed,
astronomical data, with particular emphasis on tabular data in
relational databases. An outstanding problem is how to combine
efficiently tables held in widely-separated databases; a relational
join of tables over the internet is technically difficult and wildly
inefficient.
The
preferred solution is to concentrate extracts from the original
archive tables in a single database: an astronomical data warehouse.
This allows large-scale data-mining with high efficiency. The
remaining problem is how to manage the warehouse and how to feed it
with data.
AstroGrid is developing a data grid, based on Globus and OGSA-DAI, to
feed the data warehouse. Queries on archives in the virtual
observatory will be able to stream results into the warehouse for
later data-mining. I describe the way in which grid technology has
been used for this system and the benefits and drawbacks of grid
programming.
