Measuring up the exaflood

Truly we live in troubled times. The world is running low on glaciers, virgin forest, water and, according to new research, bandwidth.

US tech analyst house Nemertes predicts we will run out of global internet capacity in as little as two to three years.

That must surely be a surprise to many Asian broadband customers. If you're reading this in Korea, Japan or Hong Kong, you may be enjoying downstream speeds of up to 100 Mbps to your home. It's hard to imagine that is going to be exhausted any time soon.

For all that, however, the report is worth reading for its methodology. 

The study points out the sheer difficulty in getting at the data needed to build forecast models.  The net is 'almost opaque' to observers, no matter how skilled they are in technical or business analysis, Nemertes says.

This is not a trivial point. Recall that ridiculous forecasts underpinned the bandwidth glut in the late '90s.  Today, for all the importance that the captains of broadband like to attach to the importance of the net, they ensure that business is conducted in the shadows under the cloak of corporate confidentiality. (And this is simple traffic measurement; we haven't even got near the even more problematic metrics for online behavior.)

That's the background. The methodology: the Nemertes study treats both demand and capacity, but separately.  The researchers decided that consumption was the metric that mattered, rather than the commonly-applied measurement of specific applications.

In particular, the research measures how user bandwidth demand has changed over time.

Choking off innovation

That amounts to a kind of Moore's Law of bandwidth demand. The more popular approach has been to mark out bandwidth actually consumed, but that can be deceiving because capacity constraints can artificially suppress demand.

Applying this formula, Nemertes' forecast is quite startling, predicting that internet access infrastructure in North America will be inadequate for demand within the next three to five years.  To bridge the gap would cost between $42 billion and $55 billion, or roughly 60%-70% more than service providers currently plan to invest.

Nemertes hastens to add that if this does come to pass, the net won't collapse. The real problem will be a choking off of innovation.

There's not a lot in these number that will enlighten Asian telcos and ISPs. The study suggests an extra $137 billion in investment would be needed to meet the global shortfall in access networking. That's a big number, and is certainly the kind of forecast vendors like to hear.

It should come as not a shock, then, that the Nemertes study has been endorsed by a vendor-heavy lobby group.  In fact the Internet Innovation Alliance (IIA), whose members include Nortel, Corning, Alcatel-Lucent and AT&T, actually co-issued the report.

Its co-chairmen, two former assistant secretarys of commerce, argued in a recent Washington Post op-ed that an 'exaflood' of bandwidth was on its way. This will require constant investment, or the country will face 'internet gridlock that cripples new services, and our country would fall further behind other countries in broadband penetration.'

 

The only way to guarantee this massive investment pipeline is simple: minimize tax and regulatory constraints and maximize competition.

Alert readers will have seen this coming. Any discussion on North American broadband demand inevitably leads back to a riff on 'net neutrality'; either an appeal for more regulation, or less. 
So it's not clear who the report was written for, and what agenda it's running to. 

It is at least an effort to forecast bandwidth takeup with a long-term demand model. For that it should be applauded. Irrespective of the methodology, in the absence of accurate data on internet traffic and underlying costs, we're all guessing.