The Weight of the Cloud
Quincy, Washington, is an agricultural city in the northeastern United States with a population of seven thousand, two supermarkets and two hardware stores. The New York Times described it as “a farming community in the middle of a desert”, and until recently its main economic activity was potato and bean harvests. But it would be no exaggeration to say that in the last few years Quincy has become one of the global capitals of the Internet. Not a capital of digital entrepreneurship, of course: Quincy has no startups and probably no coworking spaces or cafes full of people with MacBooks either. Quincy is the opposite of the flagship cities of cognitive capitalism, but online life as we know it would be unfeasible without places like this. Quincy is quite literally the Net.
Huge anonymous buildings, sometimes the size of several football fields, have been springing up in the farmland surrounding the town. Companies like Yahoo! Dell and Microsoft have picked the area as the site for several of their data centres –the big industrial facilities that are used to store and distribute all the information contained in that simultaneously personal and communal space that we now call ‘the cloud’.
The cloud, the generic name for all the services that safeguard our photos, work documents, and e-mail messages. The space that stores viral videos and hit songs, movie blockbusters and the digital maps that we rely on to find our way around cities everywhere. The space that we interact with dozens of times a day through our smartphones, tablets or PCs. ‘The cloud’ may be one of the most misleading metaphors ever coined by marketing executives, because what lies behind it is neither lightweight nor intangible. Our insatiable thirst for data has produced an enormous heavy industry that, in many respects, is not unlike the factories of the mechanical age.
Quincy is one of the key ports in the routes of global information traffic. It is joined by other similar anonymous towns such as The Dalles, Oregon; Ashburn, Virginia, and Lenoir, North Carolina, in the United States, and Sant Ghislain, Belgium, beyond it. These landmarks in the emerging geography of data are not random. In spite of Tom Friedman’s bestseller, the world is not flat, and nor is the Internet. Financial factors –land availability and tax incentives– come into play when deciding where to build them, but most importantly, data centres require direct access to infrastructures that can offer large amounts of electricity at a low cost, as well as a cold, dry climate that makes it easy to control the temperature inside the buildings, where thousands of whirring hard drives must be kept cool. Maintaining our relentless pace of information production and consumption, and ensuring that this data is available at all times, does not come cheap.
The electricity consumed by the vast industrial infrastructure that makes up the cloud is no trivial matter. Estimates of the exact amount of energy required to keep the many data centres around the world running vary, but they account for at least 1.3% of worldwide electricity use. The industry has embarked on an ongoing race to improve the efficiency of its systems, and to use renewable energy sources –nobody is keener to reduce the power bill than companies like Google, Amazon and Facebook themselves. But at the same time, the number of new facilities around the world continues to grow. And these electricity consumption figures only refer to the power used to keep the servers running. They do not include the electricity that we use to power our routers and charge our phones and computers, or the energy required to manufacture processors, hard drives, and multi-touch screens.
It becomes harder to hold onto the illusion that communication technologies operate in the virtual realm when we begin to grasp just how far they are from being a truly clean industry. The pollution emitted by data centres –particularly by the back-up diesel generators that come on in the event of a power outage, consuming fuel– are increasingly turning up on lists infringements of environmental regulations due to air pollutant emissions. For example, Amazon was cited with more than 24 air quality violations in the period from 2009 to 2011.
The reason that data centres are sometimes forced to burn diesel has to do with the sacrosanct concept of ‘uptime’, one of the cornerstones of the industry that refers to the percentage of time that a system is guaranteed to be available. Giants like Google and Facebook aspire to an uptime of 100%, and demand that it does not drop below 99.9%. In order to guarantee this level of reliability, the response capacity must have sufficient margin to keep the service available even during peaks when the demand is much higher than average. It takes many pieces working together correctly to ensure that we can upload that image to Instagram right now and not a second later, or so a new Miley Cirus or Lady Gaga video can go viral on social media.
One of the reasons why it has been easy to maintain the illusion of the intangibility of the cloud for so long is the traditional secrecy of the data centre industry, which has carried out its work discretely and remained in the background. While the seductive interactive potential of the Web 2.0, social networks and mobile internet have been on everybody’s lips over the last ten years, the technological architecture that makes them possible has developed in the shadows, without any desire to draw attention to itself. For many years, the big technology corporations have jealously guarded the number and location of their data centres. And this desire to remain invisible, which is usually based on the need to maintain strict security measures in the facilities and to prevent industrial espionage, is expressed in the design of its buildings: massive boxes lacking in architectural personality of any kind, closed in on themselves, without identifying features or corporate logos that allow passers-by to associate the container with the contents within it. Data centres may be the definitive non-places, because they are barely even buildings in the usual sense –up to 85 per cent of the construction costs of data centres goes towards their mechanical and electrical systems. And these are enveloped in as little architecture as possible.
While the internet sector has avoided talking about the industrial spaces that lurk behind its interfaces, it is still odd that there has been nothing architecturally remarkable about almost any of them. There are exceptions of course, and the most unusual of them for the last few years has been Pionen, a data centre located in an old nuclear bunker in central Stockholm that began operating in 2008. Owned by the ISP Bahnhof, it was designed to resemble a set from a 1970s James Bond film, with waterfalls, greenhouses, meeting rooms suspended in space, and even German submarine engines as decoration. In 2010, Pionen hosted the Wikileaks servers and all their secrets.
But in the last two years industry attitudes have started to show signs of change, and some experts are predicting the end of the age of anonymity. Aware of the curiosity they arouse, as the media and infrastructure tourists take an interest in their anonymous shapes, Silicon Valley giants know that these facilities are destined to become the architectural symbols of a new power; the castles of the information age.
On 6 June 2011, in his last public appearance just four months before his death, Steve Jobs showed some photos of the data centre that the company had built in Maiden, North Carolina, specifically for the launch of iCloud, the service that would store the documents of Apple users. And in December 2012, Google took the unexpected step of showing the interior of several of its data centres around the world, through a series of images by the photographer Connie Zhu.
The new data centres seem destined to accept that their symbolic nature makes them more than just run of the mill industrial infrastructure, as architects transform them into icons –into the physical embodiment of the Net. The windowless façade of the Telehouse West installation in East London, for example, is made up of rectangles that give it the pixilated effect of low-resolution computer images. Other data centres hide in full view, right in the centre of cities, cannibalising buildings that had originally been designed for different purposes. In Manhattan, the New York Telephone Company building, a grey, 32-storey skyscraper at the foot of the Brooklyn Bridge that can be seen from many parts of the city, is being gutted so that it can be turned into another container of servers and cooling systems.
While there is no end in sight to our growing data storage needs, the data centre industry continues to spread throughout the planet, beyond city centres and farmland. Facebook operates a data centre in Sweden, for example, located in the town of Lulea, just 100 km south of the Arctic Circle. And in 2008 Google filed a patent for a possible ‘floating data centre’ prototype that operates autonomously on the high seas, fuelled by the power of the wind and the movement of the waves. It is hard to tell whether this vision will end up becoming part of our infrastructural landscape, but we do know that the number of people required to keep each of these memory warehouses running –each one is currently staffed by 25 to 40 employees– is likely to keep getting smaller. The data centre of the future is an enormous dark warehouse where the light never comes on. The robots that look after their maintenance don’t need it.