Before the internet, we had the Public Switched Telephone Network (PSTN), and back in the late 1970s, it worked this way: my alcoholic Uncle Todd would, from time to time, pick up his telephone and call the telephone in my house. This created a circuit between us. The PSTN was/is a circuit-switched network. That means that my uncle’s slurred speech and off-color jibes came to my amused young ears via a dedicated pathway – a circuit between his phone and my own. The circuit lasted the duration of the call – it was born as soon as he successfully dialed (undoubtedly after a few tries) and the circuit died as soon as either of us hung up. (Usually, it was him.)
The biggest reason the internet exists today is that circuit-switching poses real problems in mass emergencies. The net is a solution to the problem that begins with the fact that my uncle and I took up one whole circuit between us the whole time during the call – nobody else could use that circuit while we were using it. He and I were tying up a whole pathway – all the wire between my house and the Joe Dano’s Bucket-O-Suds bar payphone. Since there are a limited number of pathways (pairs of wires) in the telephone network, we can’t have too many people using it at once. Reason: if everybody is using the circuit pathways at the same time, or somehow the circuit pathways are cut, new calls can’t get through.
The 1950s was the short period in US history when the Military-Industrial Complex had not yet completed the transition from its WWII war-effort origins to its current state as, uh, the current state. It was during this time when the Pentagon, through its research arm named DARPA, noticed that the PSTN and its circuit-switching had troubling implications for wartime – and wartime then meant “nuclear wartime.” The PSTN was (and still is) a huge number of wire pairs running between cities, carrying conversations on temporary point-to-point circuits. And that was the weakness of the PSTN: if the Russians were to attack the US by, say, detonating a nuclear warhead over St. Louis, point-to-point telephone communications between, say, Colorado and Washington, DC would be badly impacted, fomenting chaos. All of St. Louis’s wires would be “in use” (knocked out by the nuke) and new calls headed through there couldn’t get through there. There was more at stake than just one very weird, pork-steak-obsessed, hyper-Christian midwestern city – no less than the nation’s strategic communications were at risk. DARPA started researching the problem and ultimately the internet was the result.
The basic design goal of the internet was to replace the circuit-switched PSTN with a new packet-switched network. Unlike circuit-switching, where remember, a whole wire pair is dedicated to my uncle’s slurring, packet-switching takes advantage of digital communications technology’s ability to be instantly re-routed. All communication on the internet is, invisibly to you, broken into a very large number of very small pieces of information called packets before it is sent on its way. And for the benefit of those readers who have remained awake thus far, let us merely say that packets, unlike circuits, can and do flow around the burned-out husk of St. Louis because they can find their own way around. In packet switching, there’s no single circuit (or group of circuits) restricting flow of information. Packets are pretty cool, they brought you this page, they didn’t come to you in order, and they took a whole bunch of different routes to get to you. Packets rule, circuits suck.
So now you can imagine my nerdly shock when I checked out Nate Anderson’s piece in Ars Technica about the next-generation internet, named Internet2. Like our internet once was, Internet2 is found only on campuses, linking about 200 universities together at serious speeds.
Guess what Internet2’s newest feature is? Circuits.
The main network remains IP-based and connects more than 200 universities, in addition to limited connections to government and industry facilities. Each network segment now features a set of 10 10Gbps links, each running on a separate wavelength of light, for a total of 100Gbps of bandwidth. And that’s only the start; Internet2 says it can scale each segment to handle up to 100 wavelengths in the future. That’s… a lot of star charts.
Most intriguing is the network’s new Dynamic Circuit Network feature, which will allow researchers to set up dedicated, 10Gbps point-to-point connections across the network for short-term data transfer. The service will go live in January 2008, but it already works. In a demonstration today, Dr. Carl Lundstedt, of the University of Nebraska-Lincoln, set up a connection between his school and the Fermilab research park in Batavia, Illinois. With bandwidth provisioned, Lundstedt then transferred one-third of a terabyte of data between the two places. It took five minutes.