100G: Amazing Race in Telecom Sector

With capacity on core service provider networks doubling nearly every 12 to 18 months, the industry’s latest obsession is the evolution to 100G networks. There seem to be weekly announcements from service providers, vendors, components suppliers and industry consortiums on something related to 100G, ranging from tests to standards updates.
So the obvious assumption is that the first commercial deployment for 100G will be in the core network of AT&T, Verizon or another major service provider. That’s certainly a safe bet and likely scenario.

But I also wouldn’t be surprised if the first commercial traffic to flow over a 100G wavelength isn’t a bunch of IPTV channels or Hulu videos headed to a consumer’s home, but rather computer-assisted high-energy physics models shuttled between researchers in the United States and Europe. Or maybe it will be an algorithmic trade between a hedge fund and a stock exchange on Wall Street that is triggered by a long-range forecast in Brazil that calls for unusually heavy rainfall amounts.

It sounds counterintuitive that an enterprise or research organization would be the first to deploy a live 100G link, but it makes sense since the demand is already there.
On the research and education side, Caltech moved 1.02 petabytes of data between storage systems over a single 100G wavelength in 12 hours at a supercomputing event last fall – that’s roughly equivalent to 125,000 full-length DVDs! For Caltech and its research partners, that is critical since collaborations such as the Large Hadron Collider in Switzerland are pumping out petabytes of experiment data that need to be processed and distributed over a global grid of 150 computing and storage facilities to be analyzed by thousands of researchers. The speed of research is directly tied to the network connecting that global grid.

Or on Wall Street, where the rush to reduce latency is moving past the early adopters of quant and hedge funds to more traditional asset managers. And with some exchanges doing in excess of 130 trades per millisecond and some expecting to be pushing out a terabit of data per second in the next year, the need for 100G is obvious. Market prices may be going up or down, but trading volume is headed in only one direction – up.

In fact, supercomputing clusters, electronic trading networks and long-haul research networks were early adopters for 10G and have already dabbled in 40G, so the leap to 100G for them is clear. They have the capacity needs, technical expertise, budgets and can often move faster because these are smaller, private networks needing fewer nodes – at least when compared to a service provider’s core network.

True 100G on a single wavelength is coming, with industry standards expected in 2010 and broader commercial deployments to follow as costs come down. But don’t be surprised if before your latest video on demand purchase rides over a 100G wavelength to your TV in the next year or two that some asset manager for a mutual fund you own might make a trade over a 100G connection first.

So, who do you think will be the first to deploy 100G in a production environment? Please comment because inquiring minds want to know.