An Antidote to Data Overload

P> PASADENA, Calif. -- Those who hanker to separate cyberwheat from cyberchaff -- right now! in real time! -- may want to check out the massive parallel processing technology used by the U.S. intelligence community to filter electronic intelligence worldwide, in any language. Now, it's available to anyone with at least $30,000 to spend.

Paracel, founded by Dr. Kwang-I Yu, who developed the technology for TRW, has the financial backing of that information giant to market a hardware-software technology that can filter up to 7 million bytes per second.

Seven million per second: the power to sift through 25 novels at a tick of the clock, to find whodunit to whom and why; or, perhaps an ambitious, crafty Senate Judiciary Committee staffer late at night might put in a filter for "Foster" and "Hillary." In a couple seconds it rips through thousands of documents -- and (eureka!) the dregs of that filter may end up on front page of the New York Times.

Still, getting the Paracel technology to commercial markets has been tough -- and it's far from over. Making money in corporate markets off a language-neutral tool that gathers and sorts worldwide intelligence prompted a challenge to TRW.

"TRW and I eventually decided we could not properly market the technology within TRW," explained Yu, who holds the patent personally. So, in 1992 TRW and Yu formed Paracel, which has the exclusive license from TRW to market the technology to a world of users trying to get a grip on the Internet.

Yu, born in Hong Kong shortly after the Communist Revolution in China and educated in the United States, said the U.S. government had the problem of sifting the electronic wheat from the chaff "10 years before the Internet went into effect."

Wearing an open, short-sleeved shirt, Yu, 45, leans back calmly and confidently in his corner office and points to his computer screen. "The world comes to that spot." He pauses. "If the filter is fast enough. We have the fastest filter in the world, and our technology is inherently multilingual."

That includes filtering through the language of life itself. Yu likes to talk about the Fast Data Finder helping with the huge genome project and other biotech applications. "We have machines equipped worldwide to do gene-matching."

Yu said that with central processing unit power doubling every 18 months, the power to compute or filter the expanding pools of data must quadruple in power to keep pace.

If taming the ever-expanding Internet keeps being the thing to do, then the Fast Data Finder may become the next big thing to do it with. Yu suggested that Paracel is doing very well, but declined to cite sales figures of the closely held corporation. Paracel's Fast Information Finders cost between $30,000 and $2 million.

Here's the technical lowdown on the expensive hardware. Paracel describes its Fast Data Finder as a "search engine designed to filter, categorize and provide reports on real-time textual data such as newswires and live data feeds as well as to execute complex topic-oriented searches on massive, archived databases. A single desktop FDF-based system can categorize hundreds of megabytes of textual information across thousands of topics daily, using highly complex filtering criteria."

The heart of the technology is a system of processors that create the electronic equivalent of a garden hose. The data runs a gantlet of processors inside the "hose." The user programs, or queries, these processors to filter out particular data patterns or streams in real-time. The cheapest systems contain 3,000 processors while the high-end systems contain up to 80,000 processors.

This is serious hardware country. Using the Fast Data Finder chips for information filtering distinguishes the product from other software-based search-and-retrieval solutions from companies such as Excalibur Technologies, Verity Inc. and Personal Library Software. But that approach also makes the technology more expensive, limiting the product's target market to large government organizations and corporations -- pharmaceutical companies, intelligence agencies, brokerage houses and news services.

The Paracel home page (http://www.paracel.com) includes a demonstration that lets you check the speed of the technology.

Paracel hopes to show off its Fast Data Finder on more than its home page. Last month Paracel bought Carthage International, an information service company based in Dallas. Paracel will use Carthage to showcase the power of the massive parallel processing power of its Fast Data Finder.