REDMOND, Wash., Oct. 6, 2003 — Web services are vying for attention within the scientific community with another method for completing complex computing tasks: Grid computing.
On the surface, the two appear quite different. Web services let applications work together via the Internet, allowing — in the case of the sciences — researchers to combine and analyze data more easily. In contrast, Grid schemes link large numbers of computers to create virtual super computers capable of rapidly solving large computational problems.
But Microsoft Distinguished Engineer Jim Gray says Web services and Grid computing are synergistic. The data-centric nature of Web services complements Grid computings computational focus.
“.NET is application-centric and is designed to make it easy to build the DataGrid the integration of data sources throughout the world to produce a data library that provides a consistent, unified corpus that also is easy to query”
Gray says.
Gray points to two astronomy resources, SkyServer and SkyQuery, as prime examples of the DataGrid in action. SkyServer is a Web site developed in part by Gray and run by the Sloan Digital Sky Survey (SDSS), a collaborative research project making the most complete map of the skies visible from the Northern Hemisphere. The site offers online access to 800 gigabytes (GB) of data including images of 80 million stars and galaxies — from SDSS first public data release. In all, it presents 3 billion rows of data.
SkyQuery, another project Gray helped develop, is a Web portal that allows astronomers to cross-reference and combine data contained in disparate databases via the Internet. The portal now federates (scientific parlance for “unite”) from four online databases, including SkyServer, on two continents. As many as six additional databases are scheduled to be added in the coming months.
“The challenge,”
Gray says,
“is federating data from many archives using many different platforms by using standard protocols like XML (Extensible Markup Language), SOAP (Simple Object Access Protocol) and WSDL (Web Service Definition Language).”
Web services and the DataGrid approach conserve network costs by performing most of the computation near the data.
“It is fine to send a gigabyte of information over the network if it saves years of computation,”
Grays says.
“But it is not economic to send a kilobyte question if the answer could be computed locally in a second.”
A group at the University of Edinburgh in Scotland has done research into what they call Microsoft .NET Grid to demonstrate how .NET can simplify and enhance one component of Grid computing: Grid services, online services designed to locate the best computing resources available via the Web.
Using ASP.NET, a set of technologies in the .NET Framework for building Web applications and XML Web services, the Edinburgh researchers provide an infrastructure that offers an alternative to building Grid services with Java-based tools. In addition, .NET Grid services also allow researchers to use their Windows-based desktop computers to create Grid services that can be used to access other Grid services across many different platforms, says Edinburgh researcher Neil Chue Hong.
“By using .NET in this way, we can extend the freedom of choice and interoperability that is inherent in this technology,” Chue Hong says.
