Million-dollar customers who visit the Web page of, say, a brokerage house or an online merchant get the same response time-or delays-everybody else does. IBM, however, is putting finishing touches on a Web server management system that unabashedly separates the VIPs from the hoi polloi and can adapt itself to work on any combination of servers and operating systems.

"Businesses want to differentiate customers so they can provide preferential treatment," says Donna Dillenberger, a computer scientist at IBM's Watson Research Center in Yorktown, NY. The software developed by her group responds to changing circumstances such as the ebb and flow of Internet traffic and the kinds of requests customers make. A big order or any order from a preferred customer, for instance, is always routed around bottlenecks and breakdowns to the least-loaded servers. "A business could say that people who order a Mercedes must get the order placed within three seconds. These would be more important than people who order bicycles," says Dillenberger.

Other companies offer Web traffic management systems for specific hardware, operating systems, and applications. But IBM says its system is the first that works across all configurations, thanks to its self-learning algorithms, which let it respond to new conditions and solve new problems autonomously. The concept of preferential treatment for big customers "is one that is starting to catch on," says Don LeClair, a vice president at Computer Associates in Islandia, NY. When the IBM system reaches the market later this year, it should take a load off loaded customers' minds.