Loading ...
On Saturday I looked at Internap as a proxy for the health and growth of the wholesale IP transit market, and within that framework. Today I’m going to mention a few reasons why neither is such a great measurement for bandwidth growth as a whole.
Bypasses
A rather larger fraction of the data that end users download does not come across wholesale IP Transit backbones, it comes from caches like those of Akamai across a peering connection with each user’s ISP way out at the edge. Now, a relative growth rate for IP transit would still be comparable if the rate of growth of Akamai’s traffic were similar, but anyone checking Akamai’s growth rate over the last 5 years can see this is a really bad assumption. Because of bypasses such as Akamai, the growth of the internet at the core necessarily underestimates that at the edge. If a carrier is seeing 50-75% IP Transit growth, at the user level that may represent 75-100% growth. (Of course, this carriers is why like Level 3 and AT&T are building CDNs) But if your business sells bandwidth that comes between the internet and those such as Akamai, you are going to see rather higher growth than you would otherwise.
Geography
The internet is not ‘well mixed’. What I mean by that is that you can’t expect it to be the same at one spot as it is at another spot - even nearby. The only place the internet is ‘well mixed’ is at the wholesale IP transit layer in and between the major datacenter clusters of the world - there aren’t that many of them: NYC, Washington DC, Miami, Los Angeles, SF/San Jose in the USA, London, Paris, Frankfurt and a few others in Europe. Get outside those datacenters in those markets, and the character of the market changes, alot.
Competition
Differing geography leads to differing levels of competition. In Equinix Ashburn, for example, you can buy internet transit from any of several dozen providers. In places like Scranton PA, you can count them on the fingers of one hand and have some left over. Less competition obviously means less pricing pressure, but you also have less traffic on which to make money and therefore less benefit from scale. The economics different, sometimes very different.
So it is entirely reasonable for the wholesale IP Transit market to be in one state of health, while folks like Zayo and AFS feel much differently. If you can find the right places to be, you find greater growth rates at better pricing. Of course, there are wrong places as well, otherwise it would be too easy, eh?
5 responses to date ↓
You can follow live traffic trends at the biggest peeringpoints in the world, you will see the growth in peering is bigger in germany than in holland for example , as the DSL and cable markets evolve differently (germany gets a lot of peering from eastern europe too & holland more from the USA
Holland :
http://www.ams-ix.net/technical/stats/
Germany:
http://www.de-cix.net/content/network/statistics.html
Good post, Rob, and thanks for explaining the details. You do a great job of that.
Another reason why Internap’s wholesale IP transit business is not a great proxy for the market as a whole: the company claims they are growing quite a bit faster than the market. I’ve got a post on the topic today, here: http://ikeelliott.typepad.com/telecosm/2008/06/internaps-inter.html
Yeah I have heard them say that. I’ve seen that sort of quote before though. It is generally mathematically impossible for all the companies who say they are growing faster than the market to actually do so. It’s sort of like Garrison Keeler’s Lake Wobegone, where all the children are above average.
The public peering points can be one piece to the puzzle but in my opinion cannot be used as an accurate, direct proxy for traffic trends. The reason is that the largest carriers of bandwidth are exchanging traffic over private interconnects and not over the public exchanges.
The MINTS study is one that attempts to extrapolate Internet growth trends using data from publicly available traffic statistics (http://www.dtc.umn.edu/mints/home.html). They do a decent job outlining the challenges with their methodology as well.
Leave a Comment
Or start a new discussion in our Forum
Please consider recommending this article by clicking on the ShareThis icon!