“In small quantities data is weightless—more idea than object. But pile enough data up and it begins to fill space: closets, rooms, floors, buildings…and cities.”
A postcard grabbed from the stack at Columbia GSAPP’s Studio-X maps out six “data monuments” in Manhattan, marking the vast stores of information as bright magenta nodes in an explosive flurry of virtual information. These concentrated data centers not only take up sizable chunks of real estate (emphasis on real) but also generate enormous amounts of heat, requiring additional power—specifically half the power used to run them—for cooling purposes. Bright magenta thus seems to be a fitting color for these sweltering motherboard stockpiles.
But according to the New York Times, researchers at the University of Virginia have proposed to distribute company-owned and operated data servers into individual homes to serve as heat-generating “data furnaces,” converting an undesirable accumulation of heat into needed utilities. Any site with broadband Internet connection could theoretically serve as a micro data center, housing a concealed, slotted cabinet of servers that would produce heat to circulate throughout the building at no cost.
Inside Facebook's 147,000 sq ft data center in Prineville, Oregon, photo via.
The paper, presented in a workshop on cloud computing, calculated that the cost of building bricks-and-mortar centers to house clusters of data servers, along with energy expenses for cooling, adds up to a total of approximately $400 a year to run each individual server. What researchers propose is to decentralize the data pile-up, to reverse the concentration of physical space taken up by our virtual data, downsizing from cities and buildings all the way back to individual rooms and closets.
A data center owned by Google in Oklahoma, photo via.
According to the Times, European cities like Helsinki have already begun channeling heat generated by data centers through citywide insulated pipes for centralized “district heating.” But without such socialist infrastructure, American data centers are to instead distribute actual servers, packaged as opaque “black boxes,” to different locations while keeping them under the remote control of the company’s centralized data center. The data would be encrypted, sensors would warn if the cabinet has been opened, and if a server were to fail, it would automatically relegate its tasks to another.
Inside Facebook's data center, filtered air from overheating servers is cooled down by water sprays (right), photo via.
As seen in design solutions such as the “data furnace,” the advent of cloud computing has brought new considerations for the separation of the real from the virtual as well as public from private. To navigate this increasingly ambiguous field is to take on new ways to support an infinitely expanding “cloud” of information while responsibly sharing a shrinking supply of resources.