What will the Internet of the future be like?
Three years from now, the Net will have to carry way more data than it does today and do so more reliably.
February 3, 1999
by Angela Navarrete
(IDG) -- The Internet's about to crash. Sure, you've heard that dire prediction before, and it didn't come true. But over the next three years, the Net could face a critical shortage of bandwidth. It all begins with millions of new users flocking to the Net. According to Forrester Research, the number of online accounts in the United States alone will grow from the current 28.7 million to 77.6 million in 2002. A significant number of those accounts -- Forrester says 16 million -- will access the Net over cable or DSL connections that are potentially up to 50 times faster than today's 56-kbps modems. At the same time, many people will use the Internet for videoconferencing, telephony, telecommuting, and online gaming -- applications that not only are ravenous for raw bandwidth but demand a level of reliability the present Internet can't provide.
All of which means that three years from now the Net will have to carry way more data than it does today, and do so more reliably. This looming bandwidth crisis has the folks who run the Internet in a tizzy, from local mom-and-pop ISPs to established international telecom giants like Sprint. Players at every level of the Net are frantically scrambling to make the network, from the home desktop to the local ISP to the backbones that tie it all together, faster and more versatile. That costs money, and many of today's ISPs could be priced out of the game. Which is why three years from now, your Internet connection could be faster, your monthly Internet bills higher, and your choice of service providers more limited.
The last mile
So what do you need all the bandwidth for? Ask ISPs, telephone companies, and other Internet players, and they'll describe a typical night at home, circa 2002: Pop's in the living room, videoconferencing via laptop with his broker; while the two of them discuss the latest Internet IPO, he's simultaneously browsing the company's 3D-graphics-laden Web page. Mom's telecommuting in the den, using her company's virtual private network and virtual Centrex. Upstairs, junior's playing Quake XXII online, complete with real-time audio heckling. And everyone else on their block is doing the same thing at the same time.
You can't very well do any of that today, largely because of the bottleneck between you and your ISP -- the so-called last mile. While business users have a variety of broadband options, including T1 lines and other dedicated, high-speed connections, home users aren't so lucky: They're limited by the modem sitting on their desk. Not many home users have switched to ISDN (300,000 in the United States, according to Forrester), and at 56 kbps, today's modems are going about as fast as today's phone lines will let them.
Which is why everyone's so excited about cable and DSL, the two technologies vying to replace traditional modem connections for home users (see "Bandwidth on Demand," link below). Current cable connections can manage speeds up to 3 mbps, and that's likely to still be the case three years from now. Most DSL implementations run between 256 kbps and 1.5 mbps -- and analysts say that's what most users will get for the next three years as well.
The next stretch
But other options are on the way. In mid-1999, for example, Sprint will start rolling out its Integrated On-demand Network, or ION, which will deliver voice, video, and data at speeds up to 620 mbps. By the end of the year, businesses and consumers alike should be able to purchase an ION Integrated Service Hub at a retail store for $200 to $300, about the cost of a DSL or cable modem today. The box will be wired into your phone jack and attached to an ethernet card in your PC. Then, for prices starting around $100 a month, you'll have a persistent high-speed Internet connection, videoconferencing, local calling with Caller ID, virtually unlimited long-distance calling, and service and support from Sprint. That's not to say you'll always get all the bandwidth you pay for: Sprint admits that whenever you leave the ION network--which means anytime you ask for data from a server that's not on Sprint's network -- you'll slow down.
Two other companies, Virginia-based Teligent and New York-based Winstar, plan to traverse the last mile using wireless radio frequencies instead of congested copper lines. Voice or data would be transmitted from a dish on your roof the size of a dinner plate to a central office, which would then transmit the information to your ISP. Designed primarily for urban areas, this arrangement could be a lot cheaper than digging up the streets to lay fiber, and it would let businesses take full advantage of Net connection speeds of up to 622 mbps.
Winstar recently announced plans to expand from 30 to 60 U.S. markets between now and the end of 2000. Teligent, which will target small and midsize businesses in urban and suburban areas, entered its first 15 markets in 1998, including Chicago, Denver, Los Angeles, and Washington, D.C. The company expects to add 25 more before the end of this year.
The future of ISPs
As end users and businesses move up to faster connections, the bottleneck will shift from the desktop to the ISP.
The first problem is technical. If everyone with access to 1-mbps DSL signed up for the service, the Net infrastructure would probably be able to handle it, says Jim Southworth, director of advanced services and technology at Concentric Network. "But none of us today can sustain or support 7-megabit DSL to the home. And 20- to 25-megabit DSL will be available in a couple of years. The network will die, it will melt down, it won't handle it."
The second problem is economic. Let's say you get one of those superfast DSL services and want to view some full-motion video -- which soaks up about 6 mbps -- over the Net. Now let's say six of your neighbors decide to do the same thing. Right now, an ISP would have to pay $35,000 to $40,000 per month for a pipe big enough to handle all that data -- hardly a viable business model. Says Southworth, "DSL is going to kill the little ISP if he continues to function as he's functioning today."
As a result, it's likely we'll start seeing a hierarchy among ISPs. So-called tier-one ISPs -- the Sprints, AT&Ts, and others who have direct access to their own backbones -- will offer customers a performance advantage. "If you're connected through a tier-one ISP going to a tier-one server," says Sid Overbey, vice president of switched access and internet services at IBM, "you're going to see fantastic response time."
These same tier-one ISPs will likely position themselves as one-stop communications shops, bundling voice and data services in one handy package. For customers, the advantage would be a single monthly bill and a single number to call if any of their communications services went down. The downside: If that one company had a problem, customers could find themselves completely incommunicado.
Smaller ISPs that must rent backbone access from the tier-one companies could have a hard time keeping up. "There will certainly be a shakeout among ISPs that don't have their own backbones," says IBM's Overbey. Jim Southworth of Concentric agrees: "There are roughly 5000 ISPs in the U.S. Within the next 18 months, that number could drop by half." To survive, says Overbey, smaller ISPs will have to focus on selling value-added services such as superior customer support and Web hosting to businesses.
So what happens when millions of users connected to thousands of ISPs start exchanging ever-larger streams of data? That prospect has the folks who own the Internet backbone more than a little worried.
Internet backbone is a bit of a misnomer -- it's really backbones these days. Originally, there was just one backbone, administered by the National Science Foundation, connecting major research centers around the country at 56 kbps. But since 1995, that simple anatomy has given way to several separate backbones -- thousands of miles of extremely-high-bandwidth fiber-optic cable owned by the likes of Sprint, GTE, IBM, and MCI WorldCom, connecting major metropolitan areas all over the country. These tier-one ISPs then sell backbone access to smaller ISPs, which then sell it to yet smaller ISPs, and so on down the Internet food chain.
Right now, there's plenty of fiber. "A meteor could hit this country and bounce right off, there's so much fiber under the ground," jokes Jim Southworth. And these backbones have so much bandwidth that a lot of it is currently "dark," or unused. But what will happen when those projected 16 million cable and DSL users start swapping as much data as their fat little pipes can carry? Suddenly the fiber will light up like a Christmas tree.
That's why the backbones' current proprietors are constantly upgrading, and working on ways to milk more bandwidth from their existing fiber. The fastest backbones in use today are OC-12 lines, which transmit data at 622 mbps, but Sprint recently announced plans to upgrade to OC-48 -- 2.5 gigabits per second. "To be an Internet backbone player in a few years," says Dataquest analyst John Coons, "you'll have to run OC-48."
A bunch of aggressive new players want in on the backbone business. Take Qwest Communications. The Denver-based communications company is laying approximately 18,500 miles of Macro Capacity Fiber Network, an all-digital backbone that will connect over 130 cities by this March. Qwest plans to sell access to its network directly to businesses, offering them speedier Internet access, integrated phone and data service, and (it claims) lower costs.
But what about outlying parts of the country -- and, for that matter, the rest of the world -- that don't have easy access to this speedy fiber? The backbones connect major cities, bypassing rural areas altogether. And the undersea cables that connect continents can't keep up with today's network speeds.
Again, the answer could be in the air. Washington, D.C.-based Teledesic, for example, is currently constructing a network of 288 satellites to connect remote ISPs that can't physically (or fiscally) connect to the fiber backbone. When the satellite network is completed in 2003, an ISP in, say, Medicine Hat, Wyoming, that contracts with Teledesic could connect a small, externally mounted antenna to its servers. The dish would send the ISP's data up to one of the satellites (in low earth orbits to minimize signal delay), which would then relay the data back down to a backbone access point. Two-way connections should hit speeds of up to 64 mbps for downloads, 2 mbps for uploads. That may not be so hot compared to the rates of the U.S. backbone, but it would be a great improvement for an ISP in New Delhi whose maximum speed is now 8 mbps. "We're a niche technology -- we're not looking to take away the market for fiber," says Teledesic president Russell Daggatt. "But we'll be the only broadband technology for most of the world."
Voice and video: The new bottlenecks
You may have noticed a common thread: All these companies, from the ISPs up to the backbone proprietors, assume that three years from now, the Net will carry more than just e-mail and Web pages; it'll regularly transmit voice and video as well. But doing so requires more than adding just lots of bandwidth.
The plain old telephone system that enables you to talk to Aunt Mabel across the country relies on a process called circuit switching. When you dial out, the phone company establishes a dedicated circuit between you and Aunt Mabel for the duration of the call, ensuring that the data -- sounds translated into modulating electromagnetic waves -- all reaches its destination in one piece and in the correct order, so that Mabel hears your sentences exactly as you speak them. But circuit switching is also an inefficient use of the wire, because the circuit is held open for the duration of the call whether you're talking or not.
By contrast, networks like the Internet use packet switching, which breaks data up into packets of various sizes and sends them by whichever route has available bandwidth. These packets are then reassembled at their destination. It's more efficient than circuit switching, because if no data's coming down the pipe, no circuits are held open. Because data packets traverse the Net by whatever route is handiest, an e-mail you're sending across the state could travel around the world before it arrived at its destination. The data packets don't have to travel the same route or arrive in the right order. As long as they all get there eventually, everyone's happy.
So what happens when you try to send voice and video over the Net? You can't very well watch a movie or hold an intelligible conversation unless the packets get back and forth quickly and in the right order. As Jim Southworth puts it, "In the classic Internet model, packets could bounce from server to server and nobody would care. Real-time apps like video and voice require fewer hops."
That's where Asynchronous Transfer Mode comes in handy. ATM technology helps real-time data like voice and video traverse a packet network. First, it cuts data into uniform, 53-byte cells. Specialized routers installed along the Internet can pass these cells along much faster than it can normal IP packets, whose sizes vary. Second, ATM alters the Net's hopping-from-computer-to-computer routing system. As Southworth explains, "Even if [an ATM route] is made of many different components, it appears to the system as one hop." The upshot: ATM enables providers to guarantee transmission quality from end to end, making the Net a viable vehicle for voice and video. That's why the backbones' owners are busy installing ATM routers all along their fiber lines.
Future of the future
New cable, wireless, satellite, and routing technologies are only the beginning of the future for the Net.
For example, a select few universities and government agencies, with the help of 13 corporate partners (including Cisco, MCI, Nortel, and Qwest), are building a new Internet, dubbed (cleverly enough) Internet2. This second network will run over two backbones: Qwest's and the Very High Speed Backbone Network Service (or VBNS), a noncommercial backbone run by MCI WorldCom and the National Science Foundation -- and connect to those universities and agencies at gigapops -- network points of presence capable of transferring data at 1 gbps.
The catch: Internet2 will connect only academic and government users. The rest of us will benefit indirectly, though, as high-bandwidth Internet2 applications trickle down. "You'll see greatly improved multimedia and videoconferencing," says Vab Goel, Qwest's director of IP network engineering. For example, Internet2 engineers are developing techniques for tagging voice data packets so they get a higher priority as they pass over the network.
Another promising development is coming from a branch of physics called photonics. Right now, information travels down fiber-optic cables in the form of laser-generated light. Photonics studies ways of manipulating those light waves to increase fiber's transmission capabilities. One photonics scheme, Dense Wave Division Multiplexing (DWDM), sends multiple frequencies of light down the same fiber at the same time. Sprint's Mike Grubbs estimates that DWDM has already increased his company's fiber capacity by a factor of 32. "In the next few years," he says, "we hope to increase that to 100-fold or more." The cable ISP @Home recently signed a deal with AT&T to use the latter's DWDM-based backbone to carry its data traffic.
SilkRoad, a San Diego-based telecommunications and consulting company, has patented another photonics technology dubbed Refractive Synchronization Communication, which has demonstrated data transfer rates of 200 gbps over 200 miles, without signal amplification, on a single wavelength of light; the fastest fiber in use today maxes out at 40 gbps. SilkRoad says that it's negotiating with local telephone companies and large carriers that could roll out RSC technology by this summer. By fall it expects to offer local area network RSC-enabled hubs and switches for use by businesses and homes that have fiber connections. Some analysts predict that the entire Internet backbone will be using photonics of some sort within the next five years.
Pass the bottleneck
What you've got, then, is a game of musical bottlenecks: As one segment of the Net accelerates, the others must scramble to keep up. Right now, the bottleneck is at the desktop. As cable and DSL roll out, ISPs and backbone providers will have to speed up their segments to accommodate the new traffic. In the words of Technology Futures analyst Larry Vanston, "We'll be chasing around bottlenecks for the rest of our lives."
The question is, who's going to pay for the chase? Not surprisingly, the consensus among industry observers is that you, the end user, will -- but nobody wants to say how much. Vendors will likely try a variety of price points for broadband connections before the market settles the issue. It's probably safe to say you won't get that multimegabit-per-second connection for $19.95 a month. If users start buying all their communications services from a single company, a national case of sticker shock could ensue. "Most people don't realize how much they pay for these individual services," says Bob Hafner, Gartner Group vice president and research director, "until they see it all in one place."
You may not mind paying the price if the Internet's taking care of all your communications needs. Just pray it doesn't crash.
Networking is rife with gobbledygook. Here are some of the key concepts, in plain English.
Angela Navarrete is an associate editor for PC World.
Back to the top
© 2000 Cable News Network. All Rights Reserved.
Terms under which this service is provided to you.
Read our privacy guidelines.