Touching the internet

Journalist Andrew Blum discusses his research into the internet's physical infrastructure.

For all the profound changes the Internet has brought about, the network itself–the data centers, exchange points, and wires that make it possible–remains for most users as disembodied and non-material as the data that flows within and across it. In his recent book, Tubes: A Journey to the Center of the Internet, journalist Andrew Blum goes in search of the Internet’s infrastructure to find the physical pieces that make up our digital world. Tubes is a book about real places on the map: their sounds and smells, their storied pasts, their physical details, and the people who work there. In the following conversation with Gregory Wessner, the League’s Special Projects Director, Blum talks about what he saw on his trip to the Internet, what architects need to know about it, and what we can expect to see in the future.

GREGORY WESSER: So tell me about Tubes.

ANDREW BLUM: Tubes–A Journey to the Center of the Internet is my attempt to visit the physical infrastructure of the Internet. It comes out of about ten years of writing, mostly about architecture and buildings. What I realized around 2008/2009 was that I was supposedly writing about buildings, but I was spending all my time sitting in front of a screen all day. And at the end of the day, I would get up and look at the smaller screen that I started carrying in my pocket in 2007. There seemed to be this huge disconnect between the physical world that I was supposed to be writing about and the day-to-day life that I was living. Even stranger still was that virtual world seemed to have no physical embodiment. There was no way to bridge the gap between the world I experienced and the world on the other side of the screen. Until one day my Internet broke and a repairman came to fix it and he followed the wire from behind the couch, down to the basement, and outside to the back of the building.

GW: And you followed him?

AB: And I followed him. And then he saw a squirrel running on the wire and said, “I think a squirrel is chewing on your Internet.”  And I thought if a squirrel can chew on this piece of the Internet then there must be other physical pieces of the Internet.  So I started to wonder what would happen if you yanked the wire from the wall to see where it would lead.

GW: And so what did you find? What are the physical components of the Internet?

AB: I would say there are three different categories. First, if the Internet is a network of networks, the places where networks meet are the most important and these are called Internet exchange points. The surprising thing is that there are about a dozen buildings in the world that are far more important in order of magnitude than the next tier. So while the Internet is theoretically everywhere, it is predominantly concentrated in these dozen buildings to the point that for international communications, just as you would fly through New York or Washington on your way to Europe, your Internet traffic is always going to pass through the same key hubs, which include New York, London, Amsterdam, Frankfurt, Los Angeles, San Francisco, and others.  So that’s one category.

The second category is data centers. The generic building type of an Internet exchange point is the data center. It’s a place for equipment and things like that. But I use data centers to mean the place where data is stored, where it is warehoused, and those buildings concentrate around two poles. They stay close to where people are and/or the exchange points. Or at the most interesting scale, they’re in places that are optimized for efficiency, which at the moment means predominantly Oregon, Washington, North Carolina, Sweden, and places like that. These include the big business or big super data centers like Facebook and Google and Microsoft.  And then the third category is basically the lines between, the fiber optic cables that connect buildings, the fiber optic cables that connect cities, and the undersea cables that connect continents. Those are the three kinds of categories of physicality of the Internet.

GW: And to what degree did you see any design thinking applied to any of this?

AB: A surprising amount, or at least more than you might think.  For example, in the first category of exchange points, one of the most prominent in the U.S., and in some ways a company that’s dominant internationally, is a company called Equinix. From the outside, the typical data center–not all of them, but many of them and certainly the biggest one in Ashburn, Virginia, which has the single greatest concentration of networks in the U.S.–looks like the ultimate bland concrete box.

(Left) An Equinix Data Center in Ashburn, VA (Right) Data Center Server Room, photo: Oleksiy Mark via Shutterstock.

It looks like the back of a Walmart. There’s no sign by the door. They tell you to look for the door with the ashtray next to it to know which one the entrance is. But inside, the buildings are deliberately designed to look like you would hope the Internet would look like. They’re basically hotels and the customers are the network engineers that care for the equipment. So the interiors are meant to appeal–very explicitly meant to appeal–to the kind of sci-fi sensibility of network engineers. So that means a kind of big, red silo when you walk in. It means blue spotlights and very dramatic lighting with a sort of black ceiling like a theater. When you ask why is it dark, why are there blue lights, they pretend it’s for security, but in fact the founder is very explicit about the fact that he knew that this had to appeal to network engineers. They call the aesthetic cyberific.

GW: So a lot of these spaces are actually designed to appeal to human sensibilities and are not purely driven by technical considerations?

AB: Well, there’s certainly a technical element to the way they’re designed and Equinix prides itself in having the best technical design.  They have a patent in cable management because one element to the building’s performance is the management of the cables. It’s all about cages, it’s all about the router in a cage or a bank of routers in a cage, and then cables that go up into ceiling are strung in layers of racks. Four layers of racks, each with a different type of cable: power, fiber, copper, and inner duct, which is like super fiber. The whole building is designed to accommodate these connections between cables and that’s an explicitly designed piece.  But in terms of what the building feels like, it’s meant to appeal to a network engineers.

GW: Along this whole journey to the Internet, you met a lot of network engineers and technologists and related computer people. Did you meet any designers who had a hand in shaping any of this?  Did you search them out?

AB: The one designer I spoke with was the guy who designed Facebook’s data center in Oregon, a guy named Neil Sheehan of Sheehan Partners in Chicago. The building is a beautiful building. It has the feel of a Donald Judd sculpture to it. It comes right up out of the landscape and it’s got these beautifully clean concrete walls and a sort of light well on top with a sophisticated entry court. What’s interesting about it is that Facebook has been very explicit that this building is a showpiece. So if you believe that architecture expresses ideals, and for the most part the Internet has not had an idealistic architectural expression, here Facebook is saying with this building that they want to show the way in which they care about their practices and their infrastructure and their customers.

Facebook’s Prineville Data Center, Oregon | Sheehan Partners, Ltd. photos: Jonnu Singleton

GW: Notwithstanding this showpiece of Facebook’s, what architectural ideals has the Internet been conveying up until now?

AB: At their best, the buildings of the Internet exhibit this sort of incredibly robust functionalism that comes out of telecom, where the buildings are very thoughtfully laid out, are very strong. The way I described them at one point is they’re like very fancy plumbing supply warehouses, similar to the services provided by There’s clearly thought put into them.  They’re not just any building. Yet, they’re very deliberately anonymous and discreet.

GW: Deliberately anonymous?

AB: Because the general trend is that these are unmarked buildings. When you walk by them–and I’m thinking of one in Amsterdam that is essentially the office for one of the biggest global backbones–it’s a building that looks like it could be a mechanic shop. It’s a relatively small, relatively industrial-looking building. But if you look closely, it clearly is an expensive building. It’s clearly robustly built and built for its specific purpose. It has a very defined and incredibly functional expression. But then there is a whole class of buildings that don’t have that. They feel sloppy and fast and like basic suburban commercial/industrial boxes. It’s the same thing with the cable landing stations, the places were the trans-oceanic cables make landfall; the language they speak is of robustness and of stability. They’re clearly there to last.  But they’re also so quiet, they try not to look like anything. They are like assertively anonymous buildings.

GW: With this this new Facebook data center in Oregon, do you think that the buildings of the Internet will become more representational and less anonymous?

AB: Yes, I do. A good example is a data center in London called Telehouse West. It’s a campus with three buildings, one each from 1991, 2001, and 2011. The first one is next to a relatively significant Grimshaw building, the Financial Times printworks. The first Telehouse West building is very British High-Tech, but it was built not for the Internet but as a back office for banks. The middle one from 2001 is nothing; it is a quiet and horrible building. But the 2011 one has this very sophisticated, pixelated facade. It’s a big windowless box, but the pixelated façade is very much saying, I am a building of technology, I am a giant machine.

If the whole notion of the cloud is to replace the hard drive in the computer on your desk with a hard drive in a computer far away, as we get more sophisticated about thinking about the consequences of that replacement, you start to think more about what that far-away hard drive computer is. As soon as you start to think more about what it is, then you need pictures. And as soon as it becomes a sort of corporate emblem in some way or another then that architectural expression follows. That’s absolutely what Facebook has done.

GW: What does an architect need to think about when designing these new representational buildings of the Internet?

AB: I hadn’t thought of that. I know the trend in the past has been to express theethereality of the Internet. It’s been about fluid curves and strange shapes and somehow expressing that it’s virtual. And the future trend I think would be the opposite. It would be about expressing stability. It’s hard to say how deliberate this was but the Facebook building reads like a Greek temple. I mean, it’s got this long low building that sits in the landscape, on the top of this butte. It reads as stable, the concrete walls, the way it’s landscaped. There’s no reference to the notion that this is somehow ethereal. It feels just the opposite. But that makes sense. These buildings are the next evolution. Sort of like a bank used to be, a bank was meant to express stability. Up until now, the trust has been abstract, but I think that the trust will soon be literal and we’ll soon want to see where this stuff is and as a result, we’ll express that sense of stability.

GW: How would you say the Internet is different as an infrastructure than electricity or telephones?

AB: Well, the Internet as a whole has no designer. I mean, physically it’s the ultimate emergent system. It isn’t to say there haven’t been forces or people that made that helped define certain places. Exchange points, for example, are usually where they are for two reasons. There’s some fact of geography, like 60 Hudson Street in Manhattan is the elbow of Lower Manhattan, which has always been a communications hub. Then there’s almost always a charismatic salesman who convinced the first two networks to come and then everybody else to come on top. That’s recent, that’s in the last 15 years. But there’s always some kind of geographic fact that makes a spot important and then there’s always somebody who made it happen.

GW: So if the Internet is located where it is in part because it’s exploiting some existing infrastructure or economic system or communications history, would it look differently if you were designing it tabula rasa? Is there a theoretical ideal of the physical form of the Internet?

AB: Yeah, there is and it was the phone system. The phone system is a master planned system. But the Internet is the opposite. There is no defined structure to the Internet. It’s entirely the millions of decisions of each autonomous network. And each network is truly autonomous. That’s the fundamental idea. It’s almost a philosophical idea; it cannot be anything but emergent because it’s a network of networks. It’s always about this agreement between two networks that are in part competitors and in part cooperators.  So the transition has been from a top-down design system like the telephone, particularly the nationalized systems, to the emergent system of the Internet.  And they overlap certainly, but not exactly.

GW: How could cities be designed differently to better accommodate the Internet?

AB: There’s always the difficulty of getting fiber to buildings, but that’s a relatively small issue. The bigger issue for smaller cities is having good local hubs. Big cities don’t have that problem. In some ways their local hubs are too big. They’re too expensive. In New York the main hub buildings cost a fortune…they’re too expensive for people to get into and it pushes other people out. It decreases opportunities in some ways because of that.

But in a second or third tier city, cities do better when somebody has managed to create a building that operates as carrier neutral. It’s not owned by a Verizon or a Sprint or a Time Warner and it offers the right environment for networks to connect to each other. So you need that building for that for that connection to happen. Most cities have developed that as a matter of course, but not all. One example I looked at a little bit is South Bend, Indiana, which is lucky because it’s a railroad city. There happened to be a guy there who basically turned the railroad station into an interconnection facility. He runs it independently. And so South Bend is doing this big municipal fiber network because of the connection to the rest of the Internet that this building offers. The flip side of that, one of New York’s major interconnection facilities, 111 Eighth Avenue, which used to be privately owned and had this sort of ecosystem of different networks, is now wholly owned by Google. It is an incredible reversal of the argument that these buildings should be neutral.

GW: To backtrack to an earlier point you made about Internet concentrating in certain cities, whether it’s at these twelve major exchange points or around the larger data centers, did you then look at how those buildings are in turn affecting the neighborhoods and cities in which they are concentrated?

AB: The best example of that at moment is Oregon. Oregon’s traditional industry collapsed for the most part and the data centers are in many ways replacing that traditional industry. The data centers are encouraged because they fill a financial vacuum with the collapse of the timber and aluminum industries. They then also benefit from the power supply and the hydroelectric supply that timber and aluminum benefitted from, that infrastructure. And as enough of them settled there, then the air conditioner repair guys, the security system guys, the electricians, all the service people who grow the human capital to serve those centers set up shop there. And then more people come because those services are available. The even more dramatic example is Amsterdam. As a matter policy in the mid-90s, Amsterdam said it should be a port for the Internet as it’s always been a port for everything else. And so then said that anyone who digs a trench to lay fiber has to announce their intention. Anyone else who wants to put their fiber in that trench is allowed and they split they cost. As a result there was this incredible abundance of fiber; that both allowed Amsterdam as a city to be a key interconnection point for the Internet on a global scale. And then that also made Internet access in the Netherlands cheap and faster than anywhere else because there’s both so much fiber in the ground and because there are so many international links that wholesale bandwidth is cheaper.  So it’s just as with their airport, you can go anywhere from Amsterdam, even though Amsterdam is a city of about 1.2 million people.

GW: Did you look at how the Internet is physically manifesting itself in new cities?  Like New Songdo in South Korea or Masdar in Abu Dhabi? Is there a difference in the way the Internet evolves in a new city versus being retroactively installed in an existing city?

AB: I didn’t. But in terms of the fiber in the ground on the city scale, of course there will be a difference. It’s going to be incredibly more efficient when that piece of it is master planned. But then the connection from that city to the rest of the Internet–it either requires somebody making the expense of building a stronger direct connection, which I’m sure is what Songo did, or it will suffer from being an extra hop away and always having an extra charge of getting back to one of those hubs.

GW: What do you think is the most relevant thing for architects that you learned throughout writing this book?

AB: There’s something very relevant, which is that the hard divide between the physical and virtual worlds that we’ve been operating under and assuming for the last fifteen years can’t possibly exist. To do that is to ignore the physical manifestation of a key part of our experience. And if the connection between that physical manifestation and our experience, the thing we have on our screen, is vague for the moment, as we depend more and more on the stuff far away the connection will become more visible and more legible. An architecture magazine tweeted something like, architecture writer Andrew Blum looks at the cloud, seems a bit far-fetched for architecture. So I tweeted back that architecture is always striving for relevance. You can’t ignore the question of what is the fate of place when all of us sit in front of our screens.


Andrew Blum is a journalist writing about architecture, design, technology, urbanism, art, and travel. He is a contributing editor at Metropolis and Urban Omnibus, and his articles and essays have appeared in WiredThe New York TimesPopular Science and Architectural Record, among many other publications.

This interview took place on June 14, 2012.