Friday, October 26, 2012

OECD: Internet traffic exchange - "if it ain't broke...


...don't fix it" is the key message of a new report from the Organisation for Economic Co-operation and Development, titled Internet Traffic Exchange: Market Developments and Policy Challenges (more information on the OECD Insights blog, with coverage from GigaOM here).

The OECD report examines the history and development of interconnection and peering arrangements around the world, concluding that the current model, where agreements are on a handshake basis, with no written contract and the exchange of data happening with no money changing hands, is working extremely well and that governments and regulators should not seek to intervene in its operation.

This conclusion is very timely given the proposals to be discussed at the World Conference on International Telecommunications 2012 (WCIT-12), about which more here, particularly in relation to the consideration of whether future Internet interconnection arrangements should in future be modelled on the settlement system used for terminating international voice calls over legacy telecommunications network. The OECD's view is a categorical "no", with the report arguing that the current "handshake" model is also identifying remedies to a number of net neutrality concerns and issues.

Some key extracts from the OECD report:
"Since the Internet was commercialised in the early 1990s, it has developed an efficient market for connectivity based on voluntary contractual agreements. Operating in a highly competitive environment, largely without regulation or central organisation, the Internet model of traffic exchange has produced low prices, promoted efficiency and innovation, and attracted the investment necessary to keep pace with demand...As incumbent networks adopt IP technology, there is a risk of conflict between legacy pricing and regulatory models and the more efficient Internet model of traffic exchange. By drawing a "bright line" between the two models, regulatory authorities can ensure that the inefficiencies of traditional voice markets will not take hold on the Internet."
In the UK this is exemplified by BT's 21CN programme, to migrate its legacy networks to a digital IP network. Governments and regulators should maintain their hands-off approach to the Internet as the market has and will continue to do a good job of regulating itself:
"Evidence shows that, when allowed to do so, market participants will self-organise efficient Internet exchange points, producing Internet bandwidth to the benefit of the local economy and significantly reducing their costs, including in foreign currency. This course of action is strongly recommended in economies that do not yet have abundant domestic means of Internet bandwidth production."
The introduction of a regulatory regime based on legacy models and networks would be a step backwards, according to the report:
"A very high threshold of market failure should be established to justify intervention in the Internet market. Similarly, the growth of the Internet, together with efforts to liberalise international markets for telecommunications and promote competition, has produced enormous benefits for users, as well as for economic and social development, around the world. A treaty-based return to a regulated framework - in effect, a new settlements regime - for Internet traffic exchange would risk undoing some of those gains, while offering little potential benefit."
Also interesting is the need identified in the report to reinstate investment in basic optoelectronic physics to ensure Internet switching technology keeps pace with demand:
"An unbroken chain of basic physics research, development, and production of new technologies allowed the Internet‘s growth to keep pace with demand for the first thirty years of its existence, but investment in basic optoelectronic physics fell during the economic downturn in 2001. Consequently, the speed of network interfaces has stalled, and this has led to a transition from exponential growth to linear growth. Investment in basic research needs to be reinstated to return to a level of growth that will meet the economic and social development goals OECD countries expect of the Internet economy."
This is described in more detail later in the report:
"Historically, each new generation of optoelectronic interface has been designed to be one order of magnitude faster than its predecessor; 10-megabit interfaces gave way to 100-megabit interfaces, which were in turn replaced by 1-gigabit interfaces, and those were replaced by 10-gigabit interfaces. When each new speed of interface is introduced, it is quite expensive, but it provides new headroom. for growth. This headroom is consumed, and additional capacity is generally needed prior to the introduction of the next speed of interface, so link aggregation, or LAG, is performed, bundling two, and then three, interfaces of a given speed together, to provide some linear growth until the next order-of-magnitude faster interface becomes available and cost-effective."
However there are limits to the extent to which interfaces can be bundled together, and the  next order-of-magnitude faster interfaces are yet to become available:
"…it is not possible to simply continue adding on more switches to provide more capacity, even if it were possible to spend exponentially more money on switches, because the number of ports on each switch is limited, and switches must be interconnected with each other in a mesh, which consumes more ports for interconnection. Specifically, it requires n – 1 ports on each of n switches. A point of diminishing returns is quickly reached, at which the addition of one more switch actually decreases the amount of available bandwidth rather than increasing it…The development of switches with more ports would superficially appear to ameliorate this problem, but in fact, it merely moves the same problem from the externally-visible network topology, to the backplane of the switch, without resolving it. Only the development of faster optoelectronic interfaces, and the resultant faster network interfaces, can solve the problem and allow the Internet to resume economically-efficient growth."
We need to be making the next exponential jump to 100Gbps interfaces and beyond if the Internet is to continue to keep pace with growth and demand.

No comments:

Post a Comment