Google eyes two-story Oregon data center design to maximize efficiency

Google (s goog) is looking up for an expansion of its first wholly owned and operated data center facility, in The Dalles, Ore. Rather than sprawl out and take over lots more land along the southern bank of the Columbia River, the search behemoth plans to erect a new two-story building that will attain scale worthy of Google by standing as high as 80 feet tall, Data Center Knowledge reported Monday.

It appears that efficiency is a goal here. A document on file with the The Dalles Planning Commission about the development states that “The applicant is trying to efficiently use its property. Based on current technology, a two-story building is more efficient than a one-story building,” according to a report from the local Dalles Chronicle newspaper.

Source: Google Maps

Source: Google Maps

Stacking space vertically could let Google get more out of its square footage by cramming more servers inside it, and it’s also possible the design would require a smaller amount of fiber and power cabling, as Data Center Knowledge’s Rich Miller writes. He also noted that Google employs a two-story design at its facility in Hamina, Finland.

What’s more, building upward could mean lower total square footage, and perhaps new cooling systems could be built to fit the design. Either of these could result in less energy use for cooling. Facebook’s (s fb) data center in Prineville, Ore., achieves power-usage effectiveness (PUE) of 1.06 at its data center in Prineville, Ore., where cool air from on high gets pushed down to servers inside a tall building.

In the future, taller data centers could become more popular with the advent of robots that can quickly move up and down racks and swap out equipment.

Google, of course, is not only an infrastructure innovator, having explored floating data centers, seawater cooling and other unconventional concepts. The company has also been a pioneer of tools for using its servers to analyze great quantities of data, with MapReduce being a core example. Google Fellow Jeff Dean, a co-author of the MapReduce paper, will sit down with my colleague Derrick Harris at GigaOM’s Structure conference in San Francisco next week.