CA’s Neugents Aim to Make IT Infrastructure Smarter

Computer Associates Int’l Inc. has a concrete application in the network for its esoteric artificial intelligence technology. But now CA wants to unleash this technology into the vast stores of data in the enterprise.

CA calls its complex artificial intelligence technology by a deceptively simple name, Neugents. And Neugents is the centerpiece of CA’s Jasmine TND, an object-oriented database and development framework that was announced at Networld+Interop last month in Las Vegas. A Windows NT version is in beta testing, with final code expected later this year.

CA has high hopes for Jasmine. "We’re extremely excited about this," says Carl Hartman, CA’s vice president for information management marketing. "We believe the next wave that’s going to drive us to a $10 billion company is the Jasmine technology."

CA’s plans for Neugents correspond with the general business targets for artificial intelligence, such as sifting through massive piles of data for subtle patterns and trends. But the company first used Neugents in network operations. In April 1998, CA announced a Performance Neugent for Unicenter TNG, the company’s network management software. "It would look at the performance of a network and actually predict where and when problems might occur," Hartman says. "It would tell you, ‘In 30 minutes, your e-mail server is likely to go down.’" In December, CA began shipping a wider variety of network performance and monitoring Neugents with Unicenter TNG.

Neugents owes its heritage to the neural networks common among data mining tools. Users train neural networks with test data sets to recognize normal parameters, then let the neural network go to work on real data. Training neural networks typically takes a long time, and deciphering why a neural network arrived at a particular output is difficult because the technologies use complicated weighting patterns, not the series of either/or choices that some other forms of artificial intelligence employ. Nonetheless, once the neural network has learned the normal patterns, it can often operate very quickly on new data.

CA’s Neugents stems from the December 1997 acquisition of AI Ware Inc., a developer of intelligent decision support software using artificial intelligence. AI Ware was heavily involved in process optimization and control.

CA says Neugents is a hybrid of neural networks and other artificial intelligence methods, making the technology’s learning curve against target data much smaller.

"We expected this huge process of learning in weeks and months," Hartman says of CA’s initial tests of Neugents within Unicenter TNG. "Basically in two hours it was already telling us things we didn’t know about our network."

But outside of the relatively focused confines of network monitoring, prediction and analysis, how well will Neugents tackle less structured data across an enterprise?

"I think personally that the sky is the limit here," Hartman says. "Pick any industry. The ability to apply intelligence to the vast databases of clinical and drug data, all those things." Hartman acknowledges that while the technology is most effective when it runs against many data sources, Neugents won’t be a magic bullet. "You still need users and business people to drive what your goals are. This is the kind of technology that’s truly an enabler that business people haven’t had before."

Analyst Dave Kelly of the Hurwitz Group (www.hurwitz.com) agrees that Neugents is an interesting technology, but not magical. "It will take awhile for users to understand how to adopt it," Kelly says.

Kelly expects ISVs to be among Neugents’ early adopters. "If I’m an ISV developing a call center application on Jasmine, for example, the ability to take advantage of Neugents is going to be very helpful." Companies with well-defined, large bodies of data that may contain repeating patterns may find the technology useful right away, Kelly predicts.

"More general business use is going to take longer for CA to achieve with the Neugents technology because there’s an educational requirement," Kelly notes. "It’s not going to be a situation where you open the Jasmine box, take the Neugents and throw them at any problem you want."

Whatever the speed with which Neugents is adopted, Peter Kastner, chief research officer at the analyst firm Aberdeen Group (www.aberdeen.com), sees Neugents at the forefront of an important new trend.

"While the 1990s have been a time of enormous improvements in business productivity largely brought on by IT, I believe the next decade will see agent-driven systems increasingly monitoring the pulse of business and commerce," Kastner says. "This scenario will allow more and more workers to manage the exceptions and not the routine rules which the agents will monitor and enforce."