Tuesday, September 18, 2012

New Wave of Internet Arbitrage Coming?

The International Telecommunication Union (ITU) plans to hold a treaty conference, the World Conference on International Telecommunications, in December 2012, which will revise a 1988 treaty, the International Telecommunication Regulations (ITR). At stake are the ways communication network owners compensate each other for terminating international voice calls through the payment of settlements.

But there are wider implications. The ITU proposes to change the way the Internet is governed, in ways that will harm the Internet by raising the cost and complexity of exchanging traffic, Analysys Mason researchers argue. Basically, the ITU wants to create a new Internet traffic “settlements regime” modeled on voice precedents that will be difficult to administer and raise overhead costs.

But there could be other significant effects. First, operators might be induced to maintain their customers‘ websites abroad. One of the significant benefits of establishing an internet exchange point is to make it attractive for domestic websites to be hosted at home, in order to increase their performance and lower costs, Analysys Mason notes.

However, given that foreign websites will generate a source of incoming settlements, the incentive to keep them abroad would increase.

At the same time, foreign operators, in order to compensate for the settlements, would likely raise the price of hosting websites serving countries with high settlement rates, which might lead websites to develop less content targeted at a particular country in order to limit their costs.

While this could be seen to increase the incentives to locate content in the target country in order to avoid settlements, that is often not efficient, particularly for small or undeveloped markets from which access to a regional server may be sufficient, Analysys Mason argues.

In addition, it is likely that infrastructure investment decisions would be affected, as providers would be reluctant to invest in providing infrastructure to a particular country to which it is expensive to deliver traffic. In other words, there will be financial reasons not to build more undersea links to certain countries, for example.

Also, huge volumes of Internet traffic could be artificially generated in order to arbitrage a rate-regulated model, to generate inbound payments, alter traffic balances, or otherwise unfairly leverage any accounting rate regime that may be applied to the Internet.

Entities that believe they would be net recipients of settlements, based on current projections of traffic flows, might find themselves net payers as a result of the manipulation of traffic flows by other players.

In summary, aside from the intrinsic difficulties of successfully imposing regulations on international flows of Internet traffic, there could be unintended consequences that would harm the internet if such a system were imposed.

If history, human nature and self interest  is any guide, some service providers to try and lower settlement payments, while others will attempt to grow their share of settlement payments.

Consider the flows of traffic. The notion of settlements is that a carrier that terminates traffic incurs costs to deliver that traffic. So a sending carrier pays the terminating carrier. In many cases, the traffic flows should largely balance each other, so the net payments are relatively small in magnitude.

But there are scenarios where traffic is unbalanced, and that causes problems. In the voice settlement regime, carriers that accept more traffic than they send wind up paying money. Carriers that send more traffic than they receive make money.

Some of you will remember, or even be able to point to, instances where revenue arbitrage was possible precisely because of such asymmetrical traffic flows. Server farms and “free conference calling services” in the United States provide examples.

In the proposed ITU framework, it is server farm traffic that could be troubling for some carriers.

Multimedia content, for example, might represent as much as 98 percent of Internet traffic. Right now, where those servers are located does not have implications for inter-carrier settlements.

For cost reasons, many of those servers are located in Africa. In 1999, 70 percent of international Internet bandwidth originating in Africa went to the United States. In  2011, less than five percent goes to the United States.  

These days, content is stored at African server farms, for distribution largely to Africa, Analysys Mason notes. In some ways, that is helpful to African consumers, for quality and cost reasons. In other ways, high cross-border charges are unhelpful.

While it is true that IXPs are emerging to facilitate local exchange of traffic in Africa, the cost of cross-border connectivity between many African countries is still quite high, and this is hindering the emergence of regional IXPs to help exchange traffic and distribute content.

The bandwidth from Latin America presents the same broad picture. Between 1999 and 2011, the percentage of bandwidth going to the United States fell from just under 90 percent to 85 percent, replaced by more intra-regional traffic.

The main similarities between Africa and Latin America are that over 80 percent of their Internet bandwidth is connected to another region (Europe and the US respectively). At the same time, little bandwidth goes between countries within the region. Intra-Latin American traffic is 15 percent of total, while intra-African traffic is two percent.

The amount of cost-increasing overhead under any new settlement regime could be significant.
A recent study by the Packet-Clearing House analyzed 142,210 peering agreements representing 86 percent of global Internet carriers and 96 countries.

Only 698 of the peering agreements were based on written contracts, representing just 0.49 percent of all the contracts.

In other words, the vast majority of current international and domestic peering agreements are not just commercially negotiated, but are not even formalized in writing. If Internet settlements mirror voice practices, overall costs will rise, even if traffic flows can be accurately captured most of the time, and some say that will be very difficult.

Analysys Mason argues market-based mechanisms work, and are a better alternative to creating a new settlement regime modeled on voice principles.

No comments:

Will AI Actually Boost Productivity and Consumer Demand? Maybe Not

A recent report by PwC suggests artificial intelligence will generate $15.7 trillion in economic impact to 2030. Most of us, reading, seein...