Startup Quantiva Inc. hopes to break into Web transaction management with a tool that draws on sophisticated mathematical algorithms to automate the diagnosis of transaction problems from the customers perspective.
The Quantiva Analysis System, announced last week, is a service designed to help business-to-business site operators detect and diagnose performance problems before they affect end users, according to company officials.
The system can pinpoint early signs of problems in Web and image servers and applications, as well as determine the effects of upgrades. It can determine the effects of ISP slowdowns on customers and flag impending trouble in customer networks, officials said.
The system uses algorithms to automatically set performance thresholds after monitoring critical Web transactions and determining a traffic profile for normal patterns. When thresholds are exceeded, the analysis engine automatically runs to determine the source of the problem.
Early users at DoubleClick Inc. said they believe the tool fills a gap in performance monitoring tools, according to Mehdi Daoudi, vice president of quality of service, in New York. “There is a fundamental gap in the analysis piece. Quantiva gives you a pretty good picture of where the possible fault is,” said Daoudi.
The system uses adaptive models that refine their determination of what normal traffic patterns are for a Web transaction. “There are sophisticated statistics [gathered] over multiple time windows that work together to detect when things are normal and abnormal,” said Ron Hiller, chief technology officer of Quantiva, in Princeton, N.J.
The practice of using algorithms to take the complexity out of root-cause analysis is growing, said Jasmine Noel, an analyst at JNoel Associates, in Boston. “Vendors are trying to make it simpler to do when you have complex systems you dont want to have to write rules for,” said Noel. Mercury Interactive Corp., Altaworks Corp., Hewlett-Packard Co.s OpenView unit and ProactiveNet Inc. are all working on new root-cause analysis algorithms to automate problem resolution, Noel said.
The Quantiva Analysis System is made up of 16 data collectors placed on Tier 1 Internet backbones.
The service is available now, starting at $200 per month per URL.