Technology Advances Mean More Efficiency & Less Hardware

Key Points

• Smaller, more efficient cooling units allow more precise, flexible placement of cooling within the data center facility, reducing the mixing of hot and cold air and allowing for more effective use of limited floor space.

• More capable hardware designs and converged network architectures reduce both power and cooling needs; for example, servers that run on DC power are simpler to build and waste less energy to AC/DC conversion.

• Natural gas co-generation holds promise as both a power generation and cooling technology. Increased efficiency over conventional grid-delivered power is an added bonus.

As energy prices continue their in-evitable climb, data center managers increasingly look to improvements in power and cooling technologies and processes to hold the line on cost and risk. New approaches to delivering power and keeping temperatures in check can help IT keep everything running without risking a meltdown.

Smaller Is Better

A key trend is the move toward smaller cooling units. Traditional large, often centralized floor-based cooling devices--each of which typically takes up as much as 30 square feet of valuable floor space--are being replaced by larger numbers of smaller, more efficient cooling units. In a growing number of cases, says Jason Burnett, NeoSpire’s director of network services (www.neospire.net), units are sprouting from the ceiling.

“This approach works very well because most data centers have already moved to a hot aisle/cold aisle layout,” Burnett says. “You can put the intakes directly over the hot aisles and have them exhale air over the cold aisles. You’re not trying to cool down the hot aisles, and you don’t have to worry about exhaust heating up your cool aisles. There’s much less potential for mixing, and it’s a lot easier to direct cooling to where you need it.”

Smaller units also allow for more efficient airflow, and advanced cooling infrastructure that places the chiller in a separate unit outside the data center allows managers to avoid liquid coolant or water running directly over sensitive equipment. This more focused approach to cooling extends to the racks themselves. In-rack cooling units, where the heat is exchanged within the processor boxes themselves, can reduce efficiency losses due to ambient mixing.

“This is similar to old mainframe computers, which had in-process cooling,” says Simon Tusha, CTO of QTS (Quality Technology Services; www.qualitytech.com), a national provider of data center facilities and managed services.

Nitrogen-based cooling is another emerging technology that, unlike conventional approaches, uses a closed-loop system to remove heat. “It is more like a residential cooling system but replaces the freon that would be used in a residential system with liquid nitrogen as the coolant,” Tusha says.

Better Connections

Network architects are rationalizing their roadmaps to drive efficiencies. Asif Hazarika, senior director of product management with IP Infusion (www.ipinfusion.com), says converged networking can pay significant power and cooling dividends.

“Optimizing the use of the networks, interconnects, servers, and storage can be enabled through a converged network that reduces the equipment footprint,” Hazarika says. “Improving bandwidth by upgrading servers with 10Gb Ethernet connections, backplanes, and 40Gb Ether-net backbone and utilizing converged enhanced Ethernet and data center bridging technology will reduce power utilization and impact cooling efficiency.”

The bottom line, Hazarika adds, is reduced environmental and economic impact even as the data centers themselves continue to grow.

Smarter Servers

The trend toward reducing power and cooling at the source extends to servers, as well. Units that run on DC power are slowly making inroads into data centers. Because they don’t need to convert incoming DC power to AC, they are slightly simpler to build. Removing the transformation step results in less heat and energy loss and ultimately less load on cooling infrastructure. Still, Burnett sees a gradual takeup of DC-powered equipment.

“It does require changing the fundamental infrastructure in existing facilities,” he says. “Most data centers are set up to use AC and not DC, so it’ll take some long-term investment. Over the next 10 years, though, especially with all the green movements in place and everybody watching their carbon footprint, you’ll see a lot more big leaps in making sure servers use power in this smarter, more efficient way.”

Further power and heat reductions at the source come from later-generation virtualization-enabled server hardware. New hardware that includes built-in support for virtualization allows data center managers to take advantage of significant efficiencies not available on older hardware that wasn’t specifically built for the task.

“Three years ago, if you wanted to implement a server that could handle three virtual servers, that one machine had to have enough hardware to handle those three virtualized machines, so you didn’t gain any efficiencies,” Burnett says. “You saved space, but that was about it.”

Today’s multicore processors and faster memory allow machines to support soft CPU and RAM provisioning, which means the hardware doesn’t have to be overbuilt on a machine-by-machine basis anymore. Rationalized virtualization-specific hardware that’s matched to the application environment reduces the need to overbuild and keeps heat loads in check.

Article by Carmi Levy for .