Fast Sharing Of Huge Databases

An experimental "Web of Data" created by researchers at the University of Illinois at Chicago provides a glimpse of how businesses might one day share data over high-speed networks.

The Internet is great for sharing multimedia documents and images. But what about large amounts of data? Using the Web to browse huge, remote databases is difficult. Analyzing that data

is nearly impossible. Sending large data files over networks requires painfully slow File Transfer Protocol technology or mailing the data on tapes, says Robert Grossman, director of the university's Laboratory of Advanced Computing.

In 1999, Grossman and his fellow scientists created DataSpace, infrastructure technology for building what they call a Web of Data--an Internet for transmitting and providing access to huge databases. DataSpace consists of data-transfer protocol software, XML languages for metadata, open-source server software for making data available to users, and technology for remote data access and analysis. DataSpace runs across high-speed Sonet OC-3 and OC-12 networks.

Scientists now are deep into the next phase of the project. The Terra Mining Testbed consists of software that uses DataSpace for real-time interaction with large, complex data sets and for remote data mining and analysis. Scientists can, for example, use the testbed to access and combine data from the National Center for Atmospheric Research's database on the

La Ni--a weather phenomenon and a World Health Organization database on malaria outbreaks to study correlations between the two.

Shortly, the university will try running the testbed on an optical network called Starlight with transmission speeds up to 2 Gbps.

The project shows the possibilities of high-speed networks, Grossman says, and could "fundamentally change how we do data warehousing and distributed decision support."