Advanced Computing in the Age of AI | Thursday, April 18, 2024

The Environmental Cost of Data Snooping 

While much has been made of the privacy implications of the NSA's data profiling project, PRISM, there are also ecological impacts to consider.

While much has been made of the privacy implications of the NSA's leaked data profiling project, PRISM, a recent article in Occupy.com covers the story from a different perspective, one that is sure to capture the attention of green computing watchers.

Currently under construction, the Utah Data Center, codenamed Bumblehive, will be the largest spy center in the country when it comes online in September 2013. Costing between 1.5 and 2 billion dollars, the one million square-foot facility will house a 100,000 square-foot mission critical datacenter. The remaining 900,000 square feet will be used for technical support and administrative space.

According to former NSA analyst William Binney, the datacenter will be capable of storing "100 year's worth of the worldwide communications, phones and emails."

The article's author Peter Rugh addresses the "ill-examined ecological impacts of the site, and other smaller but similar NSA data hording facilities like it," which he says are "far from sustainable."

According to PRISM project spokesperson Vanee Vines, the warehouse-sized datacenter has been granted LEED certification from the U.S. Green Building Council, yet site plans reveal, in Rugh's words, "an extreme waste of water and electricity."

Once it's up and running, the facility will consume 65 megawatts at constant usage, costing about $40 million per year. That's roughly the same wattage required by Salt Lake City, 26 miles to the  north. Furthermore, keeping all the servers cool will require an estimated 1.5 million gallons of water a day in an area of the country where water is scarce.

Aside from the water woes of this desert topography, the Utah location is troubling for another reason. The state gets its power from the burning of fossil fuels, namely coal.

Jesse Fruhwirth, part of the group Peaceful Uprising, shares his concerns: "The 65 megawatts of energy will come from burning coal in Utah, a huge contributor to localized pollution and also to global climate change. On top of that fossil fuels are being used in conjunction with consuming life-giving water in a region where every drop is valuable."

The article also references a much-read New York Times article that depicts datacenters as massive polluters, "consum[ing] vast amounts of energy in an incongruously wasteful manner."

Those claims have since been refuted by multiple reports that show that datacenter consolidation and cloud computing lead to a net energy savings. Even Greenpeace has commended companies like Cisco and Google for their IT leadership on environmental issues. And a recent survey from the Uptime Institute revealed that the larger datacenter operators are more actively pursuing green computing strategies.

Because of a lack of data on this big data project, it's difficult to gauge exactly how efficient the Utah datacenter is. Jonathan Koomey, an energy-efficiency computing expert at Stanford University, suspects it will be more efficient that many of its private sector cousins, such as Google, eBay, etc.

"What they are probably doing is what's called 'scientific computing," says Koomey. "That's to distinguish it from the computing that a company like Google or Ebay does which varies over the course of the day depending on the demand from customers. Scientific computing applications pretty much run the computers all the time. In terms of data utilization, they're much more efficient than a lot of the data centers out in the business world."

It's worth noting that the source article is decidedly left of center, but the PRISM project has detractors and supporters of all political stripes. Whether the datacenter's environmental footprint is construed as "an extreme waste" or an appropriate use of resources will largely depend on the perceived legitimacy of the data surveillance project itself.

EnterpriseAI