Deloitte got some fantastic free press this week when they launched a their Telecommunication Predictions for 2007, and chose to summarise a fairly balanced article on transit capacity with the sentence ‘One of the key possibilities for 2007 is that the internet could be approaching its capacity’ (p4).

Bold words, that the bbc have also been publicising.
The study then clarifies on page six that what they actually mean is that demand might outstrip supply, which is very much a different thing.  Economics suggests that markets cope just fine with surges in demand by increasing the price for goods and services, so what Deloitte should have predicted is that the price of transit will rise. The degree to how much the price will change depends how elastic the demand is, or in other words, “Could we cope with not paying for as much upstream connectivity?”

I don’t think that that coping with a surge in demand should cause immediate concern, as luckily many ISPs employ clever people to think around these problems.  I’d rather see a “bad year” for the internet lead to something other than just higher prices:

• Better use of Internet Exchange points (IXPs)

More IXes, and better co-operation between exchanges in close proximity, e.g. sharing routes that they may aggregate on their route-servers.

The larger providers finding more incentive to peer widely at exchanges, to take the pressure off their (and their new peers’) congested and now expensive upstream transit connections.

• p2p software developers to take notice of proximity to peers

There’s no point growing out your peering matrix if your users still direct the majority of your traffic across the world because they have left a file-sharing tool running.  p2p software, if p2p must exist at all, needs to understand that the every user in the entire world is not your equal peer, and someone, 200+ ms away, sharing a file on the other side of the world should never be preferred over a user that you can see over peering.

• Consumer ISPs to filter traffic more agressively.

Suddenly, if the price of transit grows, it becomes much more cost-effective to cut down on unintended data use (DDOSes, virus attacks).  Consumer ISPs should operate more of a ‘walled garden’ service, proxied web and smtp access, on-network access only to DNS, maybe also access to off-network mail, ftp and VPN services, and nothing else.  This would stop many botnets in their tracks, and therefore the data that these networks produce.

• It becomes cheaper to get rid of spam instead of cope with it

A similar point to the one above; spam is now costing you a lot more money.  If you are a big ISP in the US paying a few dollars a Mbit for inbound spam from the far east, you care a lot less than a UK company that may be paying more than 20 or 30 pounds a Mbit to get the spam.  Stopping spam – an economic problem – with technology, has not stopped spam being generated.  Just because it doesn’t reach your inbox doesn’t solve the whole problem.

• Increased redundancy for transit

If transit beomes more expensive and profitable (remember, in this scenario, the price increase has been caused by demand outstripping supply, so the extra cash can be pocketed by Level(3), VSNL and the like), it becomes much easier to deliver a business case to run many more undersea cables, that follow different physical paths.  Similarly customers (ISPs in this case) will care more about availability of transit, and will therefore put increased pressure on their suppliers to improve redundancy, or take business to those providers who have done so.

So not all doom and gloom.  Yes, we may be paying a little more for our connectivity in 2008, but you could argue that consumer services are underpriced right now, killing innovation in the ISP space.  However, we may get a better service thanks to the improvements that an upward price for transit may cause.