Making the Most of Data Center Investments

By Minda Zetlin

Creating, expanding or moving a data center is a huge undertaking, and almost unthinkably expensive. No company would undertake such a move without deriving a major business benefit, right? Not necessarily.

In fact, companies sometimes create new data centers for reasons that are downright embarrassing. In at least one case, a company had to move to a new data center because the cables underlying its current data center became so voluminous that they left no room to expand.

"For years, IT staff had been simply cutting cables and leaving them under the false floor whenever they needed to change something," recalls Peter Dougherty, president and CEO of OnPATH Technologies, which provides virtualized infrastructure for data centers.

Eventually, the space between the false floor and real floor was so packed with unused cables there was no way to install anything new.

"Fixing it would have been hugely disruptive and it was actually easier to move the data center," Dougherty says. "They had painted themselves into a corner."

Avoiding problems like these means carefully planning when investing in a new data center; it also means considering big-picture issues ranging from flooring and climate control to parking and expansion capacity. Here are some key questions to answer before starting out:

1. What business benefits will the new data center provide? For C-level executives outside the technology arena, there are typically only two good reasons to create a new data center: lowering total cost of ownership or reducing long-term risk. So it's important to be prepared with detailed information about how adding or moving a facility will meet one or both of these objectives.

"More and more often, the CIO not only has to make the case for a new data center to the CFO or CEO, but also has to make a presentation at a board meeting," Dougherty notes.

In that case, he says, being able to demonstrate risk or cost reduction is even more important.

"It's often wise to bring in subject matter experts who can analyze networking infrastructure and present statistics to show what the benefits will be," he says.

2. Is one data center enough? What happens if the data center is taken off-line by fire, flood, or another local event? Many companies prepare for this scenario by keeping a separate disaster recovery center at the ready.

"My philosophy is that every company should have a disaster recovery center," says Lisa Huff, data center applications engineer at Berk-Tek, a subsidiary of Nexans and supplier of network cables. "Chances are a company will lose millions if its data center is down for a week, so it's really a good investment."

3. Outsource or build in-house? An outsourcer with data center architecture expertise may be able to build in a great deal of flexibility, Dougherty says. On the other hand, he adds, "A lot of times, there are hidden costs with a third party." This is one reason why some major companies that began with outsourced data centers are now bringing them back in-house.

Although the traditional outsourcer relationship involves both building and staffing the data center, most outsourcers will work with whatever model a customer prefers. For instance, some will deliver a "turnkey" data center, with all the equipment in place, ready to be managed by the customer's IT staff.

Huff advises keeping the main data center in-house and only outsourcing the disaster recovery site.

"And then I'd want to make sure everything was secure, and that no one else had access to my data," she says.

4. How much hardware should be virtual? "Virtualization" is a buzzword in data center design these days, and with good reason. Creating virtual equipment makes it easier to reconfigure a data center as needs and applications change, and it can allow for staff to manage center operations from remote locations.

What's more, by reducing the need for actual power-sucking hardware, virtualization can help with one of the greatest challenges for data centers today -- the use of electricity.

Power -- or lack of -- is an issue
"Electrical power, not space, is typically the limiting factor in our data centers," noted one infrastructure provider in an SEC filing that warned of the risk to its business from power supply constraints.

"Until recently, nobody cared about heating and cooling and power," Dougherty says. "Now people are seeing some real roadblocks, and there are companies on the West Coast that can't grow their data centers because they can't get access to enough power."

"The last thing any company wants to do is make the investment in a 200,000-square-foot data center and wind up using only a third of the facility because there isn't enough electricity to heat it, cool it and power the equipment," he says.

The best way to avoid this outcome is to take electricity into account when choosing the data center's location. That's what Google did when it built its new Googleplex on the banks of the Columbia River in Oregon. The river is an excellent source of both electricity and cooling.

But setting the data center next to a natural power source is far from the only option. There are many ways to reduce power needs, no matter where a data center is located, simply in the setup and choice of equipment. Consider these steps:

• Create hot and cold aisles Keeping equipment cool requires a tremendous amount of power, but alternating "hot" and "cold" aisles can help. In this configuration, rows of equipment face each other across the cold aisle, with their backs facing into the hot aisle. This takes advantage of the fact that most electronic equipment draws air in for cooling from the front, and exhausts it to the rear.

"The building's cooling system should be set up to blow into the cold aisle, and draw air out of the hot aisle," Huff says.

In this way, the equipment draws in the coolest air available, and vents the hot air where it does the least harm.

• Centralize UPS units Uninterruptible power supplies (UPS), which provide backup power to keep devices running during an outage, generally draw more power than the equipment itself, Huff says. And, she adds, most data centers have too many of them.

"The traditional setup was to put a UPS in every cabinet," she says. "But it's much more efficient to put one UPS in the center of the room for all the equipment to share, and it uses much less power."

• Shop for low-power equipment Why do data centers need so much more power than they used to? Because the hardware itself, like all things technological, is growing faster, more powerful, and smaller, all at the  same time. A smaller device using faster processors runs hotter than a  larger, slower one, and thus needs a more powerful cooling system to avoid burning itself out.

Faced with growing concerns about data center power, many  manufacturers are beginning to provide devices specifically designed to use less electricity, Huff says.

"There is equipment available now that has the same functionality and performance but lower power use," she says. "That's one obvious way to save energy and broaden one's options -- and it's a lot easier than trying to put the data center next to a waterfall."

The worry over power is a driving force behind one of the biggest trends in data center design these days: making the data center greener. "On the most superficial level 'green data center' is a marketing moniker," Dougherty notes. "But people are also using this as an impetus to get to the root of the problem and take an overall look at the true direct and indirect costs of data center initiatives. They're asking themselves: How do you build intelligently, and how do you plan for growth?"

Minda Zetlin is co-author of The Geek Gap: Why Business and Technology Professionals Don't Understand Each Other and Why They Need Each Other to Survive (Prometheus Books).