Problem solve Get help with specific problems with your technologies, process and projects.

Weighing centralized, distributed data centers

Client/servers vs. mainframes in the data center.

Mainframes have traditionally been the humming focus of data center life, even when they were surrounded by servers of all ilk. But these days, servers have enough power and speed to act as small mainframes in their own right. People talk about carving up the mainframe until it becomes nothing more than thousands of servers located in one box. Small wonder that there's also been some movement toward distributing the traditional data center out into a series of smaller centers connected by optical fiber.

"At a lot of data centers, users got impatient at the wait associated with getting information from a large mainframe, so they turned to departmental servers, which they saw as a way to get information better, faster and cheaper," says Rosemary LaChance, a principal at Farber/LaChance Inc., a data center automation consultancy in Richmond, Va. LaChance sees that trend as the first step toward distributed data centers.

But while LaChance concedes that the idea has allure, she says that most companies have realized the benefits of keeping centralized data centers -- and, indeed, are already consolidating data centers that have sprung up throughout their organizations. In the decades-long back and forth debate over centralized versus distributed IT, the pendulum seems to be swinging back toward consolidation, says Carl Claunch, a vice president and director of research at Gartner Inc., a research company based in Stamford, Conn. Claunch says, "70 to 75% of our clients are either planning for consolidation or already doing it."

That's not to say that companies should necessarily ditch the idea of a distributed data center, says Carl Olofson, program director of information management and data integration software at International Data Corp., a research company based in Framingham, Mass. But he advises organizations to first conduct a thorough analysis of their business practices. It's important that an organization's data processing scheme matches its data processing model. "The whole idea of a distributed data center is to bring the data closer to the people who need it, so that they can access it more easily," he says. "If you have a highly distributed operation, it might be useful to do this. [But] if your data needs involve keeping large amounts of data concentrated in one place because it's interdependent, it's probably more efficient to keep it on one system."

There are other factors, too, that can play for or against distributed data centers, including the following:


"Our view is that the availability of big honking bandwidth is actually going to drive more centralization than the opposite," Claunch says. The original promise of data centers was to keep data close to users as a way of avoiding expensive networking technology. "But as bandwidth gets cheaper, it's more feasible to concentrate the data and processing," Claunch says. Again, however, companies need to assess how their data interacts before running the numbers on bandwidth. "If you've got one application running in four centers -- for example, checks clearing at different centers and data must go back and forth -- yes, there would be a performance issue. But if the data is divided by geography, it doesn't matter. It can actually be a performance advantage if the data is closer to the people using it," Claunch says.


It's all about the Benjamins in this economy, and consolidating data centers can be a promising cost cutter, Claunch says. And when CIOs see that they are employing redundant people at various scattered data centers, the case for consolidation becomes ever stronger. "We've found that people contemplating data center consolidation are looking at a 20 to 25% reduction in costs, mainly from staffing," Claunch says.


There's no doubt that finding and eliminating problems is simpler in a centralized environment. Management tools are more robust, says Claunch, and it's much simpler to provide a single point for viewing the IT systems if everything is centralized.

Backup and disaster recovery

This is a double-edged sword. With everything in one place, it's easier to substitute or borrow replacement technology to solve an isolated problem, such as a server meltdown. But if a whole-site disaster such as a fire or flood were to occur, a distributed data center strategy would geographically protect data, preventing total loss.

In the end, there's no one-size-fits-all strategy. Olofson says that companies can achieve some sort of balance by mixing and matching parts of both areas. For example, it may make sense to have distributed data centers that talk to each other and send data back to a centralized data center for data reconciliation. And although data center consolidation comes with the threat of job reduction, the idea of consolidation is comfortable in the sense that most system administrators are knowledgeable about it. "Data centers were inherently centralized back in the days where companies had a few mainframes at most, so the concept is very familiar to the 390 employee base," Claunch says. "Chances are, they'll be very comfortable with it."

Carol Hildebrand is a freelance writer based in Massachusetts.


>> Visit Meta Group's "2002/03 Trends -- Enterprise Data Center Strategies" for more information on future data center trends.

>> has a variety of resources on centralized and distributed data centers.

>>Go to Gartner for detailed reports and information on centralized and distributed data centers.

Dig Deeper on Oracle data warehousing

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.