Written Presentation of
John Curran, BBN Planet, FCC Bandwidth Forum
January 23rd, 1997
Good morning. I'd like to thank the FCC for inviting me to this forum, and I will try to keep my comments brief. My name is John Curran, and I am the Chief Technical Officer for BBN Planet. BBN Planet is the Internet services division of BBN Corporation, the company that folks traditionally think of when discussing the origin of the Internet. My role at BBN Planet is to think 12 to 24 months ahead of the Internet and to try to understand the challenges we will face growing the Internet.
When I started with BBN in 1990, the Internet was a remarkably successful activity which connected several hundred organizations and several thousand researchers across the country. Over the years, we've expanded on this success to now reach hundreds of thousands of organizations and what is estimated to be in excess of fifty million people worldwide.
This growth has not been without its difficulties but due to the decentralized nature of the network, it has been possible to rapidly expand the Internet infrastructure in a very short time through independent efforts of hundreds of companies. The fact that we have been able to grow the Internet several orders of magnitude in recent years is a testament to the success of the decentralized management structure of the Internet.
By highlighting the decentralized nature of the Internet management, I do not mean to imply that the community does not work together on important issues. Nothing could be further from the truth. In important matters ranging from the scaling of Internet routing to the need for more Internet addresses, the Internet community has come together to make the changes necessary for the continued success of the Internet. The process is facilitated by a variety of forums hosted by the Commercial Internet Exchange (CIX), the Internet Society (ISOC), and other Internet associations.
Despite these successes in scaling and evolving the Internet, all is not perfect in the infrastructure. In recent months, there have been a number of reports of "network congestion" both from the user community and in the mainstream press. These reports have some basis in fact, and I'd like to outline what is being done today to address the situation. First, I'd like to spend a moment briefly reviewing the architecture of the Internet so that we have some common terminology on which to continue the discussion.
The Internet infrastructure in the U.S. is composed of nearly a dozen major backbones, each operated by an interexchange carrier (such as MCI and Sprint) or operated by an independent Internet Service Provider (such as BBN). These network backbones are composed of high-performance Internet routers and digital circuits operating at DS3 (45 Mbs) or OC3 (144 Mbs) speeds. (Note that several Internet providers have deployed digital packet or cell "switching" technologies into their infrastructure, but that is quite distinct from the voice-switching infrastructure used for standard toll calls.) The principal task handled by the Internet backbones is to take Internet Protocol (IP) data to its appropriate destination with a minimum of loss and delay. Since the Internet is composed of many different providers, it is quite likely that a given "packet" of Internet data will have to flow over multiple providers' backbones before reaching its final destination.
Connected to the backbones are customers with very high speed Internet requirements, and in addition, the networks of smaller Internet providers who often serve only a limited geography or market segment. These smaller providers may in turn service still more service providers, until the entire network looks like an inverted tree, with the available Internet bandwidth getting smaller and smaller with distance from one of the major backbones.
All of this infrastructure exists to provide service to businesses principally via dedicated leased-line circuits (e.g. 56 Kbs DDS or 1.5 Mbs T1 digital service) between the customer's local area network (LAN) and the Internet. This is the "business" Internet marketplace, not to be mistaken for the "consumer" Internet market in which customers access the Internet principally via dialup modem pools and traditional analog phone lines.
To return to the matter of Internet congestion, there are two main areas where congestion can occur in the infrastructure: in the backbones and their interconnections, and in the access network of local dialup modem pools and the associating call switching infrastructure. Congestion in the former case results in traffic loss and delay, and congestion in the latter prevents users from being able to connect due to call blockage. I'd like to focus on the backbone congestion issue for the moment before moving on to the more complex question of how to resolve congestion in the access networks.
We have evolved the Internet backbone architecture over the last five years from a model of single national backbone service, to an intermediate position of having several commercial Internet backbone providers which interconnect at key exchange points (known as network access points, or NAP's), to today's network configuration where the major backbone providers have direct interconnections between their Internet networks to ensure the fastest possible exchange of traffic. While there have certainly been problem spots in the infrastructure during these changes, I believe that we now have the necessary interprovider dialogues in place to manage interconnect capacity and am confident that the overall performance of the Internet can be scaled as needed to meet increasing traffic demands.
As we all know, a network (just as a chain) is only as strong as its weakest link. In this manner, having a robust Internet backbone infrastructure doesn't satisfy the user who can't reach the Internet due to congestion in the access network. Also, due to the inherently shared nature of the public switched infrastructure, there are other consequences of access network congestion, including blocked calls for non-Internet users.
While recent growth in the Internet has brought the issue of access network congestion into the public spotlight, the actual application of switched phone service as an access method for the Internet is no different than is done in numerous other applications, such as dialup access to corporate networks or bulletin board systems. In these applications, dialup networks are built over switched phone service, which are in turn often accessed by users at home over flat-rate residential phone service. This does bring into question some of the basic economics that apply to cost-recovery in the switching infrastructure, and the issue needs to be considered in the context of the full range of applications that can result in significant switch loading.
By its very nature, the Internet is a global construct. However, many of the most exciting activities taking place on the Internet are actually quite local in scope, whether you measure this by the user community or network traffic involved. Nowhere is this more clearly evidenced than by the abundance of city and community-based online directories which are growing rapidly in popularity. For many communities, it is now possible for one to go online and find almost everything that you'd expect in your local newspaper (e.g. local sports coverage, community events, regional calenders, and more). It is noteworthy that some of these projects are even affiliated with local newspapers (Chicago Tribune, Boston Globe, and LA Times) serving the region.
It is not just publishers that are making use of the Internet for their local distribution; we now have an increasing number of state and local governments which are "going online" to reduce costs and improve service in every area from consumer advocacy to online motor vehicle registration.
Both the community publishing and local government initiatives are good examples of the type of information services that the enhanced services exception was designed to support. While Internet service is often used for interexchange communications, it is important to recognize the rich tapestry of local information services which are being formed using the Internet as their medium. As such, elimination (for Internet service providers) of the enhanced service providers exception from access charges will have a dramatic impact on the deployment of useful information services which only now have become practical with the widespread availability of the Internet. It is hard to imagine a more effective strategy for decimating the online information industry than removal of the access charge exception for ISP's and resulting cost propagation to the end-user which will occur. If there is indeed difficulty recovering the appropriate costs for use of local switching infrastructure, then this is an issue which applies to all users of local phone service, not simply to the Internet service provider industry. Corporate dialup modem pools and bulletin board systems are but two examples of heavy usage applications outside of the Internet industry, and which suggest that local and alternative exchange carriers should have sufficient freedom in pricing of business switched lines to recover the necessary costs even in the absence of interexchange traffic calls and associated access charges. Done in this manner, any correction to the cost recovery model could distribute the costs in a more equitable manner across the entire local switched-line customer base, as opposed to singling out just the ISP industry when using such lines.
Another reason to consider directly addressing the switched line cost problem by pricing flexibility (as opposed to the indirect method of LEC's assessment of access charges) is compatibility with the open market conditions for local phone lines which will exist in many areas with the increasing implementation of the Telecomm Reform Act. A healthy market absent artificially established access charges will also bolster the development of alternative interconnect strategies such as bypassing the local switch altogether via open interconnect and adoption of new local loop technology such as satellite and cable.
In closing, it should be recognized that, while highly decentralized in its decision making process, the Internet is a very dynamic industry which will adopt new technology at a rapid pace if provided with equitable market forces. The Internet has demonstrated in the past its ability to evolve quickly to meet changing circumstances, and the current challenge to move to more efficient end-user interconnection will be resolved in a similar manner if we can avoid introducing artificial cost factors in this situation. There are very real challenges associated with appropriate cost-recovery for heavy-usage business lines, but if these concerns can be addressed through the ongoing open interconnection initiative, we have the ability to turn a problem into an opportunity which will promote the adoption of new technology and new service models.
Thank you for your consideration.
Chief Technical Officer