Bits and Bucks

The Internet has managed to hold together as it has expanded, but will continued growth upset the delicate balance of bandwidth and funding?

by Joel Snyder

Is the Internet going to collapse? Will it slowly decay over the next 6 to 12 months until it's not worth using any more? Or will the nimble technological priests and wizards who keep the packets moving continue to develop and deploy technology and resources to keep us a few steps ahead of disaster? Obviously, no one knows. In fact, not everyone agrees that a collapse is coming (see "Communications Breakdown," p. 46). It's easy to compare the Internet to a pendulum with a fairly quick cycle. A burst of usage or users or something shows up, things get bad for a few weeks/months, then someone throws a bunch of hardware or telecommunications lines at the problem and it gets better . . . until the next swing. In the past three or four years, we've seen lots of these cycles. Maybe it will go on forever that way . . .

. . . or maybe not. The cycles do have a disturbing trend. First, they are getting more and more expensive to fix. When a network access point (NAP, where the big Internet carriers meet to exchange packets) needs to be upgraded, there's a lot of money that has to change hands to boost everyone from, say, 45-Mbps lines to 155-Mbps lines. (Of course, not everyone has to upgrade at the same time.)

Second, the cycles are beginning to push more and more at the limits of the technology we have available. The Internet is the world's largest and fastest experiment in packet switching. No torture test at any one company can compare to the kind of massive volumes of data moving over the Net. When you buy nothing but top-of-the-line hardware, there's not a lot of headroom to grow. At the core of the Internet are the biggest, baddest boxes available (see "How the Internet Works," p. 54). They have room for growth: maybe even a doubling of traffic. But the way the Internet grows, that's, well, 6 to 12 months away.

The core of the Internet is balanced precariously on the talents of a small number of companies and the multitalented genii who work there. While every vendor has a fair chance to sell hardware, you see the same names popping up again and again: Cisco routers, Netscape browsers, BIND (Berkeley Internet Name Domain) name servers, just to name a few. Small defects or inefficiencies could bring it crashing down around us. On the other hand, all of this software and hardware has been tested over and over again in real-world situations and isn't likely to suddenly change.

All this means that the pendulum is going to continue swinging, and the swings will be broader than ever before. If you expect business-quality service out of the Internet, rid yourself of that notion happening in the near term. Things will definitely get better--but not before they get a lot worse.

So why can't we stabilize the pendulum and keep growth and service quality in line with each other? It's an insoluble problem because of the complexity of the Internet. There are too many factors at play. For every force dragging the Internet into the sewer, there's a counteracting force pulling it the other way. As long as those forces are in balance, the Internet will continue with periods of awfulness balanced by moments of great wonder.

No Obvious Answers
There is no physical limit to the growth or quality and usability of the Internet. Keeping the Net in running order is not a daunting task compared with problems like traveling between the stars or bringing peace to the Middle East. The answers aren't necessarily obvious, but there's no reason why a worldwide network cannot be fast, inexpensive, reliable, and secure. If we bring enough resources to bear on the problem, the outages, slowdowns, and interruptions can be kept to a minimum. But that's a big if because we have no guarantee that the awful inefficiency of the Web will be compensated forever by tremendous engineering talent.

You can understand the tradeoffs and conflicts in the Internet if you see the different forces as a series of growth curves. One curve measures the growth in services and traffic. As more and more people become connected with faster and faster technologies, the demand for Internet bandwidth grows. The traffic load increases and existing capacity is pushed to the edge. There are other compounding factors, but the key element is growth, both in demand and utilization.

The other important curve represents the profits that organizations who pay for many of the Internet services earn on their investment. Look at all the major search engines and indices as well as the hard-core information sources such as stock quotations and news. Unless they write the cost off as public relations and marketing, these companies need to make a profit to continue to fund those services. The only major search engine that isn't a commercial-laden project is Alta Vista, and that's because it's a walking, talking advertisement for the blindingly fast hardware and software that Digital produces.

Funding The Net
There are 100 different models of how to make money from--and incidentally fund--the Internet, but they all revolve around the simple fact that someone, somewhere, has to pay. A $19.95 monthly Internet access fee is barely paying for the resources you consume on a local level. Only a tiny part of it feeds the needs of the infrastructure.

It's not a question of whether the Internet needs funding; it's a question of who's going to pay it, how the money is going to be divided, how much is going to change hands, and how often. Presumably the folks who are paying the big bills see some return on their investment.

There are two key curves here: the curve in demand for services, and the curve in investment to create and deliver these services. As long as the first one stays below the second, we'll have continual balance and the Internet will speed along in the same way as my old Volvo: with occasional lurches, a bit of coaxing on a cold morning, and generally utilitarian style.

Should the folks who are paying the big bills and providing the exciting services become disenchanted with the Internet, things could get ugly and a collapse would be in the cards. We'd have an Internet where e-mail was great, but high-volume, high-content Web services were a thing of the past.

Of course, many of the companies pouring money into the Internet have no idea of what they're doing or why they're doing it, and they may go away. High-tech Ponzi schemes, poorly designed marketing campaigns, and badly set expectations will have their price, and some folks--who we probably won't miss--will leave the party.

Another reason we see so many well-placed pundits predicting the imminent demise of all that we hold dear is that the question of quality of service has never been addressed on the Internet. What amounts of throughput, uptime, and latency are considered acceptable?

For those of us who have been here for 15 years, the idea that you can transfer a 1MB file coast to coast in a few minutes is staggering. I don't care if it takes 10 seconds or 10 minutes--anything is better than waiting 24 hours for an overnight express service to deliver magnetic tapes.

On the other hand, there are those who gripe when the round-trip latency between their desk and corporate headquarters ever gets above 100 milliseconds (ms). To these users, a 350ms delay is indicative of an Internet on the verge of total collapse.

In fact, the reason the Internet pendulum moves as much as it does is because the Net is purposefully run close to the edge. That's part of the design goal of the whole system.

One reason corporate users have flocked to the Internet is that it's unbelievably cheaper than building your own wide-area network backbone. However, the reason that using the Internet is so cheap is that the service is not very fault tolerant, lines are run at maximum capacity, network operations are usually short-changed, and the network depends on good will and donated services to operate properly. The Internet could be made reliable and secure enough to bet your business on it, but then it wouldn't be the Internet--it would be something else, something much more expensive.

What can you do? Not too much. Basically, be a capitalist with an environmental leaning. Support companies that are doing things you like on the Internet and, most importantly, make sure they know you like what they're doing. Use proxy servers if available, and ask your vendor to use them. Don't be a bandwidth waster; use the "free" resources of the Internet intelligently and judiciously (see "Internet for a Small Planet," p. 76). Be a good citizen and the network will be there when you need it.

Joel Snyder is a senior partner at Opus One, an Internet consulting company in Tucson, Ariz.
Reprinted from Internet World magazine Vol. 8 No. 10, (c) 1996 Mecklermedia Corporation. All rights reserved.