Why Frankfurt, London, and Amsterdam and not as much in Bucharest, Helsinki, or Lisbon? Is it just that they got an early start with their IXPs?
Where might the next Internet hot spots emerge? Where are the next successful Internet Exchange Points likely to pop up?
I discussed your question with a bunch of colleagues attending APRICOT 2012 in New Delhi, and here are the handful of theories explaining these current Internet “hot spots” along with their corresponding predictions for the future.
1) The Healthy Internet Peering Ecosystem theory suggests that exchange points emerge and thrive in an ecosystem where there is a high concentration of eyeball and content traffic. Eyeballs love content and vice versa, so colocating them increases the gravitational pull for other eyeballs and content. The eyeball networks and content companies might then all peer with each other to bypass the metered upstream transit service (see “the business case for peering” for the financial proof). Once there is enough local traffic, the international ISPs, CDNs, etc. become interested in the IXP as a landing point for the region and for that local eyeball and content traffic. The gravitational pull then kicks in and other international ISPs want to build in to pick up that local and international traffic.
This theory predicts that a healthy and growing Internet region ecosystem (demonstrated by broadband penetration growth, dropping transit costs, a growing and cooperatively inclined Tier 2 ISP population, dropping prices for transport, etc.) will eventually have a thriving peering population that could become a powerful regional IXP.
2) The Cable Landing/Fiber Nexus Theory suggests that locations that are topologically close to the most expensive part of the Internet service –the transoceanic transport capacity to the U.S. for example – will become thriving Internet peering ecosystems, because, much like the well-suited ocean harbors became the centers of regional trade in the olden days, these drop points become the Internet centers for exchange. Even better, places where multiple cables land and/or cross would make building into an emerging IXP less expensive. The larger ISPs then purchase their own transoceanic capacity for their Internet services. This concentration of carriers and ISPs leads to a concentration of Internet transit providers who are trying to POP close to the lowest cost access to the Internet on the other end of that fiber.
This theory predicts that the next hop spots for the Internet might be the drop points for the new cable systems.
3) The Geographic Proximity Theory suggests that the desirability of London, Amsterdam and Frankfurt are in part due to what one can do once they get there. For example, some believe that London would be a good first European POP for later expansion across Europe. Frankfurt is seen as a good landing place for ISPs from the Middle East and Eastern Europe. The value of the IXP includes its proximity to the next expansion sites. Conversely, Australia is on the way to nowhere, so a network extended would have to be justified solely from the benefits of exchange traffic with the locals there. There is a good peering population there, but the geographic proximity theory suggests there is little chance of it becoming a regional IX location.
This theory suggest perhaps Shanghai and Hong Kong would be a next hot spot because of its proximity to the billions of people in China. Locations in the Middle East and Africa would also fare well here.
4) The Financial Center Theory, proposed by Arnold Nipper, suggests that the financial capital markets are an IXP accelerant. The financial community has always paid to drop milliseconds of latency and this in turn helps drive carriers and ISPs to service them topologically close to the financial districts.
This theory explains why London, Frankfurt, Amsterdam, New York, Chicago and Tokyo are all Internet hot spots in the U.S: they all have thriving financial exchanges as well.
This theory predicts that the next Internet hot spots would be places like Dubai, Milan, Shanghai, etc. as the side effect of servicing these financial sectors accelerates these sites to the next level of regional peering.
5) The Business-friendly theory was put forward by Matt Moyle-Croft and suggests that an unstable legal and difficult regulatory environment will slow any attempt to build and evangelize a regional IXP. Examples include convoluted or onerous regulations, unpredictability, unexpected application of rules, ISP licenses required for international ISPs, local business entity requirements, etc.).
This theory predicts that the business-friendly and loosely regulated environments in the U.S. and Europe will enable thriving IXPs, while there would be few IXes across the more turbulent countries in the Middle East and parts of Africa for example. Just about any government peering initiatives such as India’s NIXI will scare off international ISPs or content providers looking for a regional POP.
So if we blend these theories together, we can rank the emerging markets and make a guesstimate as to where the next hot spots will be. For a student/intern project, it would be an interesting exercise to take a look at a list of IXPs, and rank them along these dimensions to see if there is a correlation between the success of the IXPs and their relative strengths in these areas.
Thanks to Arnold Nipper and Matt Moyle-Croft for allowing me to share these discussions to share.