Coordinating global internet traffic

Net traffic is doubling every hundred days, and electronic commerce should reach $300 billion by 2002.

Some parts of the communication industry are asking governments to address an increasing number of regulatory issues. Local telecommunication carriers are seeking redress for inequities they believe are arising from the final delivery of Internet traffic for dial-up users in respect to regulation aimed at telephony. Some smaller Internet service providers (ISPs) claim that arrangements for traffic exchange with larger ISPs are not transparent and non-discriminatory. Some international facilities-based ISPs argue that they are not being fairly compensated for carrying traffic for other parties. Finally, some ISPs are calling for professional management of the Internet's global root name servers that provide crucial information for the Domain Name System (DNS). All these issues will require decisions by policy makers, even if those decisions ultimately give governments a minimal, but supportive, role in finding solutions, for example by providing a neutral forum for industry to reach consensus or by providing a neutral point at which statistics can be aggregated to inform debate.

The Internet is turning traditional point-to-point communication models on their head. For example, in the world of public switched telecommunication networks (PSTN), there was generally a correlation between the amount of traffic exchanged between two countries and the amount of capacity allocated by infrastructure providers for this route. For the Internet, the points of origin and termination of packets of data have very little relationship either to the traffic that may be carried over a given route or to traditional PSTN traffic patterns. For example, PSTN traffic between Singapore and Canada makes up well under one per cent of each country's outgoing international traffic. In December 1997, by way of contrast, nearly half of the international Internet capacity from Singapore's Internet exchange point (STIX) was deployed between Singapore and Vancouver. Similarly, in the PSTN world, telephone numbers were administered on a national basis. For the Internet, more than two-thirds of second-level domain names registered by Canadian, French and Spanish users are administered in the United States.

The Internet has evolved in a fundamentally different way from the world's PSTNs. National PSTNs were established locally and connected internationally many decades later. The reverse is largely true for the Internet. Until relatively recently, an e-mail message between two users located in the same building, but with different ISPs, was likely to travel over inter-continental backbone networks and receive DNS information from a server located on the other side of the globe. In some countries, Internet performance is better on an "international" than on a "national" or "local" basis, and in a number cases, this is directly related to past, and present, regulatory obstacles to infrastructure competition.

Due to the Internet's evolution, as well as the price and availability of national and regional infrastructure, a great deal of traffic between users in one geographical area is being pushed onto international or intercontinental infrastructures, particularly where this improves performance. However, this may negatively affect performance for users in other countries by creating congestion in traffic exchange (including domestic traffic exchange) in those countries. Local infrastructure performance is no longer isolated from local and regional performance elsewhere in the world. Consequently, governments need to increase the momentum for liberalising infrastructure provision, notably by implementing the World Trade Organization (WTO) agreement, not least because poorer levels of performance in the local infrastructures of other countries affect their own markets.

The private sector, and in particular ISPs, are actively developing infrastructure to "localise" Internet traffic flows by improving the performance of local infrastructure. Their initiatives include the establishment of an increasing number of Internet exchange points (the places where ISPs exchange traffic) and greater global distribution of the infrastructure supporting the DNS. In these cases "localisation" does not indicate a specific geographic area but rather the fact that content, services, and some network functions are being shifted closer to the user to increase network efficiency (e.g. to bring users faster response times). These developments are critical in improving the performance of the Internet for electronic commerce. This document aims to increase awareness of how traffic exchange occurs on the Internet and how infrastructure and financial interconnection arrangements (including peering and transit) to accomplish traffic exchange are developing. The role of government in these developments is to eliminate existing and potential barriers by monitoring and reforming regulation as appropriate.

Because of the international nature of the Internet and the rapid pace at which the services and infrastructure are developing, some of these issues are complex. Moreover, the nature of traffic exchange on the Internet, relative to the PSTN, does not easily lend itself to past commercial settlement models. For example, if the traffic flowing between a country in the Asia/Pacific region and the United States were contained within a single international link and within the geographical borders of both countries, a model for sharing infrastructure costs might be self-evident. However, the Internet does not transport traffic in such a precisely defined or bounded way. Not only might the IP (Internet Protocol) packets travel along different paths and through different countries, so might the DNS information drawn from global root servers.

From the perspective of ISPs in the United States, the greater the capacity foreign ISPs put in place to Internet exchange points based in the United States, the greater their costs in providing domestic infrastructure that is used for international transit. From the perspective of ISPs outside the United States, the existing peering model (akin to "sender keeps all" in the PSTN world) does not fairly compensate them for costs incurred in carrying the traffic of US users. There is little information available for understanding the balance of these costs, or even the patterns of international traffic (including transit traffic and traffic related to international DNS requests), although such information is necessary to inform discussion of the financing of international infrastructure linking OECD countries. At this stage, the best way forward is for industry to initiate discussion on the financing of Internet traffic exchange, for example via the Asia Pacific Internet Association's call for comments and other industry forums. The role of government is to stay abreast of these discussions and support industry-led solutions.