Technology becomes crucial for data mining
Scientists at the Stanford Linear Accelerator Center (SLAC) at Stanford University in Menlo Park, Calif., are studying the smallest things to try to make sense of the biggest. Their careful, meticulous analysis of subatomic particles could one day explain why there is a universe at all and not just a sprawling expanse of nothing.
In a sense, grappling with such enormous ideas is just part of what high- energy physicists do every day. At the Stanford center, which is funded by the Energy Department, the experiments produce enormous amounts of data, which the scientists store in an object database made by Objectivity Inc., based in Mountain View, Calif.
Object databases vie with relational databases in the market and each have strengths and weaknesses.
Relational databases store information in tables, the virtual equivalent of a spreadsheet. They are good for storing simple data such as the prices of products or billing data. Object databases, a companion to object-oriented programming languages such as C++, store data as discrete objects with numerous interconnections between one piece of data and others.
Object databases are ideal for large volumes of data, especially when scientists or investigators need to discover relationships that are not obvious, said Jay Jarrell, Objectivity's president and chief executive officer. When data is stored in relational databases, it must be broken into pieces — such as the name of a terrorism suspect and the terrorist events he is suspected of — with each data point stored in its own field and linkages coded between them.
Object databases can create a discrete "object" that contains information about the suspect and the incidents. "We do not have to decompose those objects that are coming in in massive quantities. We can store them directly," Jarrell said. "It allows the researcher the capability of finding elusive needles in a sea of haystacks that are distributed logically and physically around the world."
"We can take any object that can be described in [an object-oriented programming language] and put them into our database, no matter how complex," said Leon Guzenda, Objectivity's chief technology officer. "This data is complex because it has many relationships with other objects."
"We've been seeking ways to handle our information for decades, and we've always been putting our toes into database technology," said Richard Mount, director of SLAC Computing Services and assistant director of SLAC's Research Division. "In the early 1980s, I was sticking my toes into relational database technology, and it was obvious it was a lousy match. Our data didn't fit the relational model very well."
Mount began considering object databases, and Objectivity specifically, in the mid-1990s, he said. "Many years ago, it was clear that Objectivity had a product that was able to scale to what we needed. In the mid-1990s, as we looked much more deeply, it became apparent there were things that would need to be fixed, but there were no showstoppers."
The laboratory began using the database in 1999 and now stores more than a petabyte of data in it, he said.
SLAC is one of Objectivity's more public government customers, but it is not the company's only one. The company counts the Air Force, the Navy, the Defense Intelligence Agency, the National Imagery and Mapping Agency and several other agencies among its customers.
The object model is not universally accepted, so Objectivity is trying to convince more agencies to try it.
"When you start to develop these complex applications that are object oriented, when they start trying to stick the objects into three-dimensional rows and tables [of a relational database], there's an abstract [mapping] layer they have to write," Jarrell said. "It costs the government, it costs the integrator, it costs everybody about 40 percent more. The integrator can charge more body shop money, but a lot of times they cannot solve the problems, and it will fail."
The concept is still relatively new and perceived as unproven, though that may be an unfair assessment, according to Dave Capka, chief scientist and a technical fellow at Northrop Grumman Mission Systems.
"Just about everyone has somebody who's experimenting with object databases," he said. "People still do see them as something of a boutique technology. I have come to conclude that this is a real technology, but that view still isn't shared by everyone in the community."
SLAC's BaBar sitehttp://www-public.slac.stanford.edu/babar/
"Lightning to strike Los Alamos" [Federal Computer Week, August 25, 2003]/fcw/articles/2003/0825/tec-lightning-08-25-03.asp
NEXT STORY: Oracle still on PeopleSoft trail