April 1998

BY BROUGH TURNER

The Internet is experiencing explosive growth, but will it — can it — continue? And if so, how will this growth be funded? Before proceeding, let’s touch on a more basic question: What exactly is the Internet? Today, the Internet is an interconnected mesh of thousands of independently managed networks. It includes perhaps 5,000 distinct service provider networks and an untold number of private corporate networks providing service to tens of millions of people. This mesh is in a constant state of flux because of continuous growth and the autonomous actions of its numerous participants.

The Internet can best be viewed as a multi-tiered entity, comprising public Network Access Points (NAPs), national backbone carriers, regional carriers, local and national Internet Service Providers (ISPs), corporate and institutional networks, and individual users. Activity at each level impacts how the Internet as a whole evolves. And, as we’ll see, new tiers, overlays, and specialty service categories are emerging.

Typically, entities at one level buy Internet connectivity from service providers at a higher level. At the top level, there are competing providers of Internet backbone service. Before the National Science Foundation shut down its original Internet backbone in 1995, they defined a set of interchange points, the NAPs mentioned above, where commercial backbone providers exchange messages destined for computers that are connected to their competitors’ backbones. The summer 1996 issue of the Boardwatch Directory of Internet Service Providers listed 9 national backbones. By the fall issue there were 14. As this is written, there are 40 national backbone networks identified at www.boardwatch. com/isp/backbone.html.

BACKBONE CONSOLIDATION — WHAT?
Two years ago, many people were predicting an imminent shakeout and massive consolidation among ISPs. There have certainly been acquisitions, but the total number of ISPs continues to grow — to over 4,500 at the latest count. Today’s predictions are for consolidation in the Internet backbone, with one or a few players emerging with near monopoly control. Attention is focused on WorldCom. With the UUNET backbone and extensive fiber lines, WorldCom was already a major Internet backbone provider by the end of 1996. Then in 1997, WorldCom successfully acquired the backbone networks of Compuserve and AOL (while giving Compuserve’s retail customers to AOL as part of a three-way deal). With WorldCom’s impending purchase of MCI, commentators woke up and started predicting drastic consolidation in the backbone — even monopoly control by WorldCom/UUNET.

But in any growing market or industry, real business consolidation doesn’t actually occur until growth rates slow. As long as the rate of growth holds up, individual acquisitions are offset by the emergence of new players and new approaches to the market.

Indeed, while a WorldCom/MCI merger might carry close to 50 percent of today’s Internet backbone traffic, new players are rapidly emerging, many with new approaches. Qwest is laying enormous quantities of fiber and leasing it to any backbone provider they can find (AGIS, CRL Networks,…). In addition, they have announced voice-over-IP in head-to-head competition with traditional long-distance carriers. Sprint, with roughly 23,000 miles of fiber in the ground since the mid-80s, is rapidly retrofitting their network with Wavelength Division Multiplexing (WDM) equipment. Their current WDM equipment multiplies fiber capacity by a factor of 16 — the equivalent of another 350,000 miles of channel capacity added to Sprint’s original network — without a backhoe! Who knows what AT&T has in the ground? Or when their new management will figure out what to do with it! Looking forward three to six years, there are some enormous potential networks in the offing, including multiple competing satellite systems and global fiber schemes like Project Oxygen.

At the same time, totally new approaches to improved backbone performance are emerging. For example, caching schemes leverage the low cost of disk storage to store copies of popular pages close to the customer, reducing delay and saving backbone bandwidth.

A private NAP strategy is exemplified by SAVVIS. Instead of going through the public NAPs, SAVVIS is willing to buy high-capacity access to their three largest competitors at multiple points (the SAVVIS private NAPs). Typical backbone providers carry traffic to the nearest interchange point and hand it off. SAVVIS carries traffic as far as possible on its own network and only hands off at the last moment through systems they control. The result: SAVVIS provides substantially better performance to their corporate customers. Indeed, SAVVIS has consistently ranked first in independent surveys. With this performance advantage, SAVVIS is serious competition, even to a UUNET-MCI combination.

ACTIVITY AT THE ISP LEVEL
There are only 25 or 30 million people connected so far. And, connecting new users to the Internet remains a complex task, with widely differing requirements for home users, SOHO, and small, medium, or large businesses. None of today’s national carriers have customer support centers up to this task. Local ISPs do. Not only that, but local ISPs are targeting niche markets — lawyers, contractors, businesses serving college campuses, and so on. So, local ISPs will remain a growth market for some time to come.

THE IMPACT OF END USER ACCESS
One circumstance that helps competition at the local ISP level is that there is currently no natural monopoly in the technology used to provide end users with Internet access, namely modem calls over the telephone network. ISPs are competing by offering customers dial-up service, with all ISPs having equal access to dial tone. This could change as customers demand higher capacities. The new 56 Kbps modems operate right at the limit of what can be sent over a 64 Kbps telephone circuit. For higher capacity one must have access to both ends of the wires — be it phone lines and xDSL modems, or cable and cable modems.

As discussed in my On The Horizon column in October 1997, xDSL uses advanced modem electronics at each end of an existing copper pair to increase its capacity from 56 Kbps into the 1 to 8 Mbps range, depending on the condition and length of the copper wire. Globally, the telephone companies have a window of opportunity to reuse their wire, but whether or not this will happen is an open question. If xDSL deployment is left to the monopoly telcos, then it’s likely to be deployed as rapidly as ISDN was, that is, hardly at all. If third parties get access to the bare copper wires, at plausible prices, then substantial xDSL deployments are more likely.

Right now, cable modems are pulling ahead in the race to provide increased bandwidth to the home. Cable modems take advantage of the existing coaxial cable used for cable TV. The original cable TV networks were one-way only — broadcasting TV towards the home. Making a two-way infrastructure requires major upgrades to the existing cable networks, above and beyond adding cable modems in each home that subscribes. Eighteen months ago, these upgrades looked unaffordable for the debt-laden cable industry. Now, with the promise of a new market, and aided by more than a billion dollars of investment by Bill Gates, cable company stocks are up, and their ability to finance the needed upgrades is dramatically improved.

What’s the point of all this? When the home user’s bandwidth demand goes beyond 56-Kbps service into the realm of xDSL or cable modems, market dynamics will change. We could see a decrease in the total number of ISPs, depending upon who gets to control the wires to individual homes. This would not happen overnight, and with wireless access on the horizon, it may not happen at all. So for at least the next two to three years, we can expect a competitive environment — with many local ISPs, improving service, and competitive rates.

FUNDING THE GROWTH OF THE INTERNET
Some futurists project bandwidth that is "too cheap to meter." Nice idea — but it won’t happen anytime soon. For the next couple of decades, there are enough potential new applications to consume all the bandwidth being discussed. After all, I don’t yet have 3-D holographic virtual presence conferencing with my co-workers in Europe, or with my sister on the other side of the country. And we can’t avoid paying to fund the investments that provide the bandwidth and connectivity needed for today’s exciting applications and the killer applications of tomorrow.

On the other hand, there are those who expect to charge per-call for Internet telephony. In the long term they’re kidding themselves. With the Internet, the marginal cost of delivering another packet is zero, until you reach the point of network congestion. Attempts to provide premium services on a call-by-call basis (using RSVP for example) haven’t scaled well, so attention is shifting to schemes that provide a few basic service levels (using precedence bits in the IP header).

With these approaches, it is easy to monitor the average and peak flow rates for one, or a few, different types of packets as long as you don’t have to measure any one subscriber for more than a tiny percentage of the time. The peak rate is determined by your connectivity — 28.8 Kbps or 56 Kbps dial-up, T1, etc. The average flow rate can be monitored with a brief measurement every five or ten minutes. So, in the future, we will likely see a continuation of today’s flat monthly charge for raw access based on the kind of connectivity you sign up for. In addition, there will be a higher-tier charge for priority packets (say for high-quality voice telephony). This charge will be a flat rate or will be based upon the average of many brief measurements, widely spaced in time. In a competitive market, it will not be based on counting individual packets or telephone calls.

WHAT ABOUT VOICE TELEPHONY?
In 1996, data bits exceeded "voice" bits, in the United States. Five years from now, at current growth rates, the bits required for all the voice conversations in the world will be a trivial percentage of the bits being passed over the Internet. All that’s required to precipitate dramatic restructuring of the traditional telephone network is an appropriate grade of packet service on significant parts of the public Internet. When that appears, watch out. Until then, watch what is happening with voice over private IP networks.

CONCLUSION
Will the exponential growth of the Internet continue for decades? Yes! The underlying technology, both silicon and fiber, continues to improve. And demand for bandwidth will continue to grow, driven for now by Web applications, but soon by voice and video communications, new forms of broadcasting and totally new applications. Moore’s Law was about a semiconductor technology, but there should be a similar rule for human potential. Humans will continue to invent technology that, among other things, increases available bandwidth. And humans will find innovative ways to use all available bandwidth. For those new services that provide value, customers will pay for bandwidth, fueling the continued growth of the Internet.

Brough Turner is senior vice president of technology at Natural MicroSystems, a leading provider of hardware and software technologies for developers of high-value telecommunications solutions. For more information, call Natural MicroSystems at 508-620-9300 or visit the company’s Web site at www.nmss.com. E-mail to the author ([email protected]) is also welcome.