(Image of data center via Google)
Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway.
– Andrew Tanenbaum, Computer Networks
Michelle Brumfield, a rural resident of Yorkshire, was tired of slow Internet speeds. So she carried out a stunt. Brumfield uploaded a video to Youtube, which a BBC reporter 120 kilometers away waited to download. At the same time, a partner attached a memory card that held the same video to the leg of a racing pigeon. When she hit “Upload,” her partner released the pigeon.
Which made it to the reporter first? The pigeon. In the 75 minutes it took for the pigeon to fly 120 kilometers, only 24% of Brumfield’s video was uploaded.
That’s unfortunate, you say, but thank goodness that we here in America don’t have to rely on the indignities of offline delivery networks when we want Youtube. But wait. We’re here to tell you that a form of that system exists in America too. The world’s biggest content delivery networks (CDNs) all rely on airplanes to make their sites run smoothly. Yes, airplanes. Your sleek, modern Internet experience relies more on offline data delivery than you think.
No, pigeons aren’t powering the Internet. But the online world as you know it is made possible in part by Boeing jets, pallets, and cardboard boxes. Here’s why.
It’s easy to ignore the fact that the Internet relies on a vast physical infrastructure made up of servers, data centers, and fiber optic cables. But it does, and where you access your data from matters. People in Australia, Finland, and South Africa would have a much harder time streaming music and photos if everything came out of a single data center located say in Utah.
0.06671281903: That’s how many seconds it takes for light to travel to the other side of the planet. It seems small, but it’s significant. When you add in the relays and switches that the data has to go through on the way there, it adds up.
That’s why we have CDNs located all over the world. CDNs are systems of servers that host data at local internet service providers (ISPs). To have your modern Internet browsing experience, you need a local CDN. It’s thanks to them that your movie doesn’t have to buffer during a car chase scene.
Here’s a concrete illustration of how CDNs work. In 2007, a CDN installed smart servers in thousands of Starbucks coffee shops. These servers contained some of the most popular songs on iTunes. Customers waiting in line for coffee found that the normal download time for a song was cut from 30 seconds to two seconds.
Not every website has to set up a server in more than one location. But high-traffic ones do, or their load times would be unbearably slow. It simply takes too long to send large data files over the Internet. Dan Rayburn, an authority on media streaming technologies, has made the calculations: “Even if you have a private connection from your office to a data center, transferring 1000 movies can take up to a month.”
So instead of using the Internet to transfer big pieces of data, companies have turned to the global freight network. High-traffic websites copy data onto hard drives (which are no bigger than what you’d use to back up your laptop), pack them into cardboard boxes, and then fly them around the world. They can be in a box in the belly of a passenger plane, right beside cartons full of iPhones.
Let’s say you want to send a big file to a friend nearby. Rather than emailing it, it might take less time if you download it onto a hard drive and then drive it over. That’s a calculation that high-traffic websites have to make when they want to move massive files out to the edges of their CDN, and the number of websites relying on freight is higher than you think.
The Internet, a series of tubes after all. Image via Google.
To better understand why the world works this way, it helps to dive in to the physics.
In spite of rapid improvements in computing power, bandwidth capacity—the amount of data that can be sent over an Internet connection—has not kept up. We’re all familiar with Moore’s Law, which predicts that computing power grows at an annualized rate of 60% a year, implying a rate of doubling of every two years. You should know about two other, less famous laws. The first is Kryder’s Law, which maps the growth of data storage; the second is Nielsen’s Law, which maps the growth of Internet bandwidth.
Dr. Mark Kryder, who formulated the eponymous law, explained to us in an email that Moore’s Law and Kryder’s Law have advanced at roughly the same rate. That should make sense. “Computer architects would say that it’s desirable,” he remarks. “If the speed of a computer didn’t increase with its storage capacity, you wouldn’t be able to do much with the stored data.”
But increases in bandwidth capacity have not kept up with Moore’s Law and Kryder’s Law. While storage capacity and computing power have grown at 60% a year, growth in bandwidth capacity clocked in at just 50%. That is, hard drives are getting smaller—and therefore cheaper to transport—at a faster rate than improvements in Internet bandwidth. So companies have been finding it cheaper, and faster, to send data by air or sea in hard drives rather than over the Internet.
Most interestingly, given the delta in the rates of growth of these laws, we should expect more and more Internet traffic to be delivered offline. The more data we’re able to store in a given physical size, the faster it is to fly those files. Counterintuitively, technological progress is actually favoring moving more data by jets, not less.
But it’s not just cutting edge CDNs that leverage airplanes to move bits around the world. The Internet in Cuba is extremely limited, due not just to government censorship but also poorly-developed infrastructure. Still, ordinary Cubans have found ways to get their digital fix of Game of Thrones, Taylor Swift, and the New York Times. How? Simple: someone takes it there by plane.
Instead of subscribing directly to Netflix or Spotify, Cubans pay a company called “El Paquete Semanal.” That translates to “The Weekly Package,” and it (probably) works like this: Every week a person flies from Miami to Havana, bringing hard drives loaded with last week’s movies, music, and articles. El Paquete copies these onto smaller hard drives, which it then delivers to subscribers. It’s so effective that Vox has referred to the organization as “human data traffickers.” Though El Paquete can’t recreate certain online experiences, it is effectively hand-delivering the Internet to Cuba.
That’s just another way that offline technologies deliver online experiences.
What are the implications when bandwidth doesn’t grow as quickly as storage and computing power? Computer scientist Jim Gray made this assessment in 2003: “We have an embarrassment of riches in that we’re able to store more than we can access. The fundamental problem is that we are building a larger reservoir with more or less the same diameter pipe coming out of it. We have a much harder time accessing things inside the reservoir.”
That’s why people still slam their fists on the table waiting for their websites to load, and why downloading movies still takes so long. And companies will continue to ship data around the world as long as growth in bandwidth capacity lags behind growth in storage and computing.
Consider that on your next trip abroad, there may be a better selection of movies traveling in the belly of the plane than on the seatback in front of you.