Replace your data warehouse

If your data warehouse is more than five years old, you're going to need a new one, according to a Gartner analyst.

Speaking at the Teradata Partners conference in Las Vegas, Gartner analyst Donald Feinberg told attendees they face massive changes to data warehousing.

"I'm going to tell you why you're going to have to redesign your whole data warehouse if it's five years old," Feinberg told the audience.

He said people call him up and complain: "My data warehouse I built five years ago or ten years ago doesn't work anymore." Such complaints stem from changes in industry and IT making older data warehouses unable to keep up.

Data warehouses are becoming mission critical. "Mission critical means if data warehouse goes down for more than an hour, you're going to have to go to manual systems," he explained.

The amount of information being used is exploding, he said. "The real problem is with the internet. Nobody expected data warehousing to be used as repository for online transactions," he said. "It's a whole different set of rules."

As business intelligence becomes more pervasive, all that information will need to be accessed by even more users, including those outside a company, such as suppliers, partners, distributors and even customers.

"Everything is going to be designed around the worker," he said. "Five years from now, more people working in your business are going to be from the Information Age - grown up using computers all their life." This will affect the whole way they do things, he said: "Do you think you're going to be able to tell them which programme to use?" Such a shift means IT will need to create services and let workers put together how they'd best like to work, which will change the demands placed on data warehousing.

Feinberg stressed that throwing more hardware at data warehouses will not solve the problem. Rather, firms must optimise, model and design them properly.

Properly designing data warehouses will keep companies from having to replace them again in five years time, said Feinberg. "We need to figure out how to design data warehouses to be flexible today so we don't have to change over again."

He added that the major changes in data warehousing are happening now, and future evolution will be less drastic. "If we do it right now, there will be no major changes in the future," Feinberg said.