I've had the pleasure of attending the APRICOT 2013 conference in Singapore this week. As always, it has featured an interesting technical agenda — part emerging standards and technologies, and part operator experience discussions. Many of the faces are familiar — the technology developers show up in different parts of the world to talk through recent IETF working group items, etc. And, they take away valuable insights about how local networks and businesses work, that could impact how a technology will deploy. It turns out that, while Internet technologies are global, impacts and evolution are not always uniform.
Even more interesting to me are discussions that focus on an analysis of "the Internet in the wild" — real world data about the Internet collated and brought for consideration among operators in an open forum such as this. This is a key feature of the Internet: it is accessible; anyone can join it, study it, and discuss it. Contrast that with a traditional telephony network, where the only folks with insight into how a network is functioning is the operator of the network. (See our whitepaper "Internet Invariants: What Really Matters" for more musings on the invariant principles that make the Internet all that it is).
I was particularly struck by Kurtis Lindqvist's presentation in the first session on Tuesday morning: http://www.apricot2013.net/program/apops-2 . Kurtis' presentation was on "The History of Peering in Europe and What this Can Teach Us About the Future". As advertised, he stepped through some of how the Internet evolved from the early academic days through to modern times. The reality of the "early days" was that much of the traffic from Europe was destined for the US, where most of the content of interest was. Accordingly, network links were built to accommodate the traffic. As time went by, more local content was developed (local newspapers, etc), and it became more interesting to form local peering agreements to make more efficient paths between clients and servers of that local content.
More time goes by, and we begin to see the "mega content" providers – according to some estimates, Google (in all of its properties, including YouTube) accounts for up to 12% of traffic between (a customer) ISP and "a large peer". The rest of Kurtis' talk was about whether this is history repeating itself, and if yes, what might we do about it now?
You can (and should!) go look at his presentation or the video archive to see the pretty graphical illustration of these points (and the full version of his suggestions for more CDN and peering build out), but for my purposes I want to stop and look at Kurtis' presentation at APRICOT from the outside in.
Kurtis pulled together his data from available studies, drew the conclusions, and put the illustration out for the room full of network operators and engineers to consider in the light of "where do we want to go with this?" The Internet is a network of networks — what the Internet becomes depends on what these operators (and others around the globe) will choose to do to address this kind of situation.
But being able to articulate this perspective and back it up with concrete data in an open meeting is one thing that makes the Internet fairly unique in terms of key commercial infrastructure: it is accessible, it can and is discussed, and collaboration is required to find the best path forward for one and all.
This is truly governance of the Internet — note the little "g". It is discussions like these that govern what operators decide to do to improve their networks and the Internet as a whole. And it's very refreshing to watch in action.