40Gbits Granny and the Future of Telecoms

The news that Sigrid Löthberg, Peter Löthberg's 75-year old grandmother, has the world's fastest Internet connection has now passed through the Web's digestive tract. All the oohing and aahing is complete. It's certainly very cool that she has 40Gbits/s connectivity and a CRS-1 router in her garden shed, but it is only a demonstration project.

However, it does tell us quite a lot about how Cisco thinks the future will be. And there will not be a CRS-1 in every pot any time soon. The fibre, well, that's a different story. Sweden, like some other countries, has a number of projects that aim at the creation of shared, open-access fibre infrastructure. It's a question of getting the institutions and economics right; aggregating enough customers to spread the capex while guaranteeing open access to preserve competition. And that, by the way, is what the Digital Town strand of Telco 2.0 is all about, and the focus of the Digital Cities summit on the 18th of October: find out more here.

Cisco's website has all the specs you could ever need on the CRS-1. It's pure engineer porn; an IP router with a switching capacity of 320Gbits/s in its smallest configuration, with 40GBits/s linecards, using native IP directly over DWDM fibre. Fibre wavelengths can be used up to 2000 kilometres from the CRS-1 without regeneration. And, should this somehow prove insufficient for one's requirements, more CRS-1 units can be added to the system without going offline, up to a total of 92 terabits/second. Whoo.

Who might ever need such a thing? The clue is in the name; "CRS-1" stands for Carrier Routing System One. Nobody who isn't a telco, except perhaps the NSA or the Chinese secret police, is ever likely to require one of these. When the first ones appeared, in 2004, it was said that the figure 1 stood for one CRS for each country. But there's a long tradition of these predictions proving wrong; IBM boss Thomas Watson, of course, said there was a world market for maybe five computers. What people forget about that remark is that, when he said it, "computer" meant something even more technologically exotic, and expensive, than a CRS-1.

More of them than expected have been sold, to some surprising customers. Mobile network operators, facing unexpectedly high demand for data after deploying HSPA, have been notable buyers, usually around the same time they decide to pull fibre to their heaviest cell-sites. To condense, though, the CRS-1 is proof that Cisco expects telcolike organisations to exist, and to depend entirely on IP, and access network bandwidth to rise enough to fill the machine. Condensing even more, and exposing it to Telco 2.0, someone's got to handle the bits, whatever else happens. It's also worth pointing out that Cisco is deeply uninterested in whether telcos do IMS, pure SIP, or IP-SDP; for them, the next-generation network is IP.

Even if you're a networkless, pure application service provider, you're still going to be routing eyewatering quantities of bits if you're doing it at telco scale - so you're a potential customer. Even if you're Skype, your success depends on the bit pipes having sufficiently powerful pumps. The lesson from Google - and Vodafone, which decided to invest heavily in core networking from the beginning - is that you can't fake infrastructure. If your applications are of equal quality, the side with the best infrastructure wins. If your apps are of somewhat lesser quality, and your infrastructure better than the competition, you still might win. You can pretty-up dodgy applications; you can win through marketing; but if the infrastructure can't take it, you're doomed.