Illustration by Terry Widener

24 n W

The Dawn of the

“Stupid Network”

FeaturedStory

Mixed Medi a

DAVID S . ISENBERG

i“Stupid Network”

IN RECENT HISTORY, THE BASIS OF TELEPHONE

COMPANY VALUE HAS BEEN THE SHARING OF

SCARCE RESOURCES — WIRES, SWITCHES, ETC. –

TO CREATE PREMIUM-PRICED SERVICES. OVER THE

LAST FEW YEARS, GLASS FIBERS HAVE GOTTEN

CLEARER, LASERS ARE FASTER AND CHEAPER, AND

PROCESSORS HAVE BECOME MANY ORDERS OF

MAGNITUDE MORE CAPABLE AND AVAILABLE. IN

OTHER WORDS, THE SCARCITY ASSUMPTION HAS

DISAPPEARED, WHICH POSES A CHALLENGE TO THE

TELCOS’ “INTELLIGENT NETWORK” MODEL. A

NEW TYPE OF OPEN, FLEXIBLE COMMUNICATIONS

INFRASTRUCTURE, THE “STUPID NETWORK,” IS

POISED TO DELIVER INCREASED USER CONTROL,

MORE INNOVATION, AND GREATER VALUE.

26

David S. Isenberg <isen@isen.com> is the principal “prosultant” of isen.com (http://www.isen.com), a

business he founded in January to rethink the value of telecommunications networks from first principles.

Telephone companies have always pushedtechnology improvements that promote thesmooth continuation of their basic business.They invented the stored program controlswitch in the 1970s, as a move toward costreduction and reliability. Programmabilityalso made possible certain call routing andbilling services. In the 1980s, phone compa-nies began marketing these services as the“Intelligent Network.” Technology contin-ued its trajectory of improvement, butbecause technology began to change thevalue proposition in ways that the old busi-ness could not assimilate, the telcos seemedto fall asleep at the switch at the core of theirnetwork. Meanwhile, the Stupid Network —based on abundant, high-performance ele-ments that emphasized transmission overswitching, as well as user control of the vastprocessing power at the network’s edges —was taking shape.

KEEP IT SIMPLE, STUPID“Keep it simple, stupid,”or KISS, is an engineeringvirtue. The IntelligentNetwork, however, is any-

thing but simple; it is a marketing concept forscarce, complicated, high-priced services, sur-rounded by features like 800 service, call wait-ing, and automatic calling card validation.These are intertwined in the network architec-ture in a plethora of service adjuncts, eachwith its own systems for operations, provi-sioning and maintenance. This complicated,centrally controlled amalgam of systems isborn of a single application — two-way real-time voice communication.

So what exactly is a Stupid Network?George Gilder observed more than five yearsago, “In a world of dumb terminals and tele-

phones, networks had to be smart. But in aworld of smart terminals, networks have to bedumb.” In a Stupid Network, control passesfrom the center to the edge, from the telco tousers with an abundance of processing powerat their fingertips. The center of the network isbased on plentiful infrastructure — cheapbandwidth and switching — that is about assmart as a river. The water in a river, like adata object in a Stupid Network, gets to whereit must go adaptively, with no intelligence andno features, using self-organizing engineeringprinciples, at virtually no cost. Bits go in oneend and come out the other. Data flows — likewater — define the movements and channelswithin the system.

Eric Clemons, a professor at theUniversity of Pennsylvania’s Wharton Schoolof Business, makes the distinction betweenstrategy and doctrine. “Strategy,” he says, “islearning how to deal with dogs. Doctrine isabout belief: ‘Dogs don’t do that.’” Telcodoctrine, formed in the age of monopoly andscarce infrastructure, is rarely examinedexplicitly. When there was only one tele-phone company, what Ma Bell did definedhow things were. So today, even though thenew era of competition requires clear think-ing and new beliefs, telco culture inextricablymixes doctrine and strategy.

THE CIRCUIT-SWITCHED LEGACY The Intelligent Networkconcept has its roots in thefirst software-controlledswitches in the 1970s. In

those days, working with computers meantwriting code to save a byte here and aninstruction cycle there. Current software prac-tices, such as object-oriented programming,were too inefficient for prime time and were

MIXED MEDIA FEBRUARY/MARCH 1998n W

27

for revenue-producing services geared towardcall set-up and billing. The concept of networkcontrol was extended to let digital switchescommunicate with databases (known asService Control Points) and signal processingsystems (Intelligent Peripherals).

Intelligent Network specs were meant toencourage telecom equipment vendors todesign their equipment to work in a multi-ven-dor environment, so telcos would not belocked into one supplier. In addition, thisequipment was designed to work with certaincustomer systems and databases. Some com-mon Intelligent Network services include:routing calls to a number other than the onethe caller originally dialed (the basis of 800service); caller options (“press 1 for customerservice,” etc.); and supplying calling partynumbers directly to customers for databaselookup (which is why I must call my bankfrom my home phone when my new ATMcard arrives in the mail).

STUPID IS BETTERStupid Networks havethree basic advantagesover Intelligent Networks— abundant infrastruc-ture; underspecification;

and a universal way of dealing with underlyingnetwork details, thanks to IP (InternetProtocol), which was designed as an “internet-working” protocol. Some key “two-fers”emerge from these basics: Users gain end-to-end control of interactions, which liberateslarge amounts of innovative energy; innova-tive applications are rapidly tested in the mar-ketplace; and innovative companies attractmore capital and bright people.

Abundant InfrastructureIn a Stupid Network, if you have congestion,you just add more connections, bandwidth,switching or processing power. If you wantreliability, you add more routes or moreredundancy. If you need more intelligence for

relegated to the confines of academia.Thus, telephone network equipment was

designed in a climate of scarcity. Considerthe local exchange, represented by the threedigits of a telephone number that follow thearea code (the nxx in the pattern nxn-nxx-xxxx). The local exchange “owns” the lastfour digits of a telephone number.Theoretically, a local exchange can serve upto 10,000 telephones, e.g., with numbers762-0000 through 762-9999. The designassumption, though, is that only a certainpercentage of these lines, maybe one in 10,are active at any one time. But should theseassumptions change temporarily (e.g., anearthquake in California) or permanently(calls to AOL lasting several times longerthan normal voice calls), the network hits itslimit. Then, getting a dial tone becomes amatter of try, try again.

Even more assumptions have now changedpermanently. Before 1996, long distance car-riers like AT&T, MCI and Sprint cranked aprecise set of capacity planning equations thattold them which switches would need morecircuits, which routes needed to be upgradedover the next year, and where to plan for newswitches. Suddenly, increased Internet usagethrew the telcos an unplanned 60% increasein data traffic. Suddenly, some points of theirnetwork hit capacity. The telco fallback posi-tion had always been to lease capacity fromtheir rivals, but this wasn’t available either,because the other telcos were maxed out aswell.

“INTELLIGENT” NETWORK SERVICESIn the late 1970s, telcosbecame fixated on theirexpensive investments incompute r- con t ro l l ed

switching, and were intrigued by the prospectthat they could do “intelligent” things withthese investments. They reduced the cost ofrunning the network and formed a platform

MIXED MEDIA FEBRUARY/MARCH 1998n W

Mixed Medi a

Stupid Networks

have three basic

advantages over

Intelligent

Networks —

abundant

infrastructure,

underspecification

and

internetworking.

features or services, you add it at the end-points. As early packet network visionaryPaul Baran points out, it is possible “to buildextremely reliable communications links outof low-cost unreliable links, even links sounreliable as to be unusable in present networks.”

Even as the costs of networks havedropped, capacity has improved manyfold. Atthe dawn of the digital transmission era, forexample, you could run 1.5 megabits — 24calls — on a coaxial cable as thick as yourankle. Today, network providers routinely putseveral tens of gigabits — a few hundred

thousand calls — on a single glass fiber as thinas a human hair. Switching used to be scarce,too, but now it is equally abundant. Where ahuman operator could set up maybe 100 callsan hour, modern computer controlled switch-es, such as Lucent’s 4ESS, can now completeabout 1 million calls in the same hour.Furthermore, if you consider that routing asingle packet is equivalent to setting up a call,routers can now set up 3.6 trillion “calls” anhour. And the prices for these componentshave come way down: Today when you buy aGigabit Ethernet switch, you get 1,000chunks of 64 kbps throughput (each equiva-lent to a phone call) for every dollar.

This leads to two different models of capi-tal investment. In the telco model, networkexpansion is a big decision that requires expertengineers, detailed Erlang models of conges-tion, and several consultants — and takesmonths, if not years, to implement. But ifyou’re running a Stupid Network, expansionis no problem. A few hundred gigabits? Put iton my credit card. More switching capacity?Take it out of petty cash.

UnderspecificationThe Intelligent Network is tightly specifiedfor voice. All other data types require specialleased access lines, or awful kludges likemodems. The Stupid Network is underspeci-fied — this means bits-in, bits-out. It is noth-ing special for underspecified networks tocarry voice, music, bank balances, e-mail orTV on the same facilities. You stuff bits inone end of the network, and they find theirway to the other end of the network. Packetscarry their address with them, and out theycome at the other end, right where you wantthem to be.

Underspecification also means that thereis little thought for congestion control. Sowhat if there is congestion, or even crashes?On the whole, the convenience of underspec-ification more than makes up for the occa-sional jam-up. And you can always add more“infrastructure.”

InternetworkingThe Internet Protocol points the way to a keyproperty of Stupid Networks. The foremostdesign goal of IP is to cross multiple, physical-ly different networks. To IP, it doesn’t matter ifthe underlying transport is circuit, SONET,Ethernet, Bitnet, FDDI or smoke signals. An IPapplication works the same no matter whatthe underlying network technology. Thismakes the details of how a network worksirrelevant (including how “intelligently” it isengineered).

IP neatly takes the provider of the physicalnetwork infrastructure out of the value propo-sition. No matter how intelligent a telco’s net-work might be, if it is running IP, itsintelligence is reduced to commodity connec-

MIXED MEDIA FEBRUARY/MARCH 1998n W28

If you’re running a Stupid Network, expansion is no problem.

A few hundred gigabits? Put it on my credit card. More

switching capacity? Take it out of petty cash.

29MIXED MEDIA FEBRUARY/MARCH 1998n W

Mixed Medi ativity. Networks that run IP are left with onemain source of distinction: how much connec-tivity they provide. Thus, the Internet that weknow and love is a “virtual network” — a“network of networks” — that is independentof wires and transport protocols.

Because IP makes the details of the networkirrelevant, all that matters is that the bits sentby your machine are received by my machine,and vice versa. In an IP communication appli-cation, users don’t care how the “call” is setup, or even if there is a telephone call thatforms part of the communication pathbetween endpoints.

This means that users are in control oftheir interactions. Suppose, for example, thattwo users want to bring a third party into aninteraction; they just do it. An IP-connecteduser does not need to order special three-wayconnectivity service from the networkingcompany. All that user needs to do is write (orinstall, or use) a program that sends packetsto two different destinations and receivesfrom both of them.

A BOOST TO INNOVATIONThis ability to “just do it”liberates huge amounts ofinnovative energy. If I havea Stupid Network and I get

an idea for a communications application, Ijust write it. Then I send it to my buddy, andmy buddy can install it, too. If we both like it,we can send it to more people. If people reallylike it, then maybe we can charge for it — oreven start our own company.

In contrast, the Intelligent Networkimpedes innovation. Existing features are inte-grally spaghetti-coded into the guts of the net-work, and new features must intertwine withthe old. For example, until recently you couldnot get Caller ID for an incoming call whenyou were on the phone. To fix this, Bellcorehad to invent a low-tech, low-functionality,high-complexity protocol called Analog

Display Services Interface (ADSI). Call waitingwith caller ID, however, would be a no-brain-er under Internet telephony — a packet con-taining caller ID information could be treatedjust like any other packet arriving at my sys-tem, to be interpreted by the end-user IP tele-phony application.

New Internet capabilities are comingonline that will fuel Stupid Network innova-tion. Internet Protocol Version 6 (IPv6), stabi-lized in 1995, is becoming available now. Itscapabilities include expanded address space,real-time functionality, mobility managementand carrier selection, hooks for authenticationand data integrity, multicasting, and easycoexistence with and migration from the cur-rent IPv4 standard. IPv6 capabilities in thehands of innovators will foster whole newareas of applications.

QWEST FOR BANDWIDTHPerhaps the leading propo-nent of the Stupid Networkis Qwest Communications,which is building a huge

16,000-mile-plus SONET-based network thatwill reach 125 cities by next year. According toQwest’s executive VP of products, NayelShafei, “Qwest is leveraging the greatest cur-rency of all — our unlimited bandwidth — toshape the future of telecommunications.” Thecompany is about to go international, too; itwill turn on its 1,400-mile Mexican backbonein mid-1998, and it recently announced atrans-Atlantic link.

The Qwest network will run native IPover SONET, according to CEO JoeNacchio. This is an industry trend; Sprint, forexample, recently went ATM-less on itsSONET/IP backbone. This is possiblebecause physical layer infrastructure isbecoming more abundant and endpoints arebecoming more capable.

Qwest is using its network to competewith the telcos on price. Qwest now offers

New Internet

capabilities,

such as IPv6,

will fuel

Stupid

Network

innovation.

30

long distance telephone service over its IPbackbone at 7.5 cents per minute, a 25%discount over the now industry-standarddime.

Shafei maintains that the sound of tele-phone service over Qwest’s network will nothave the glitchy, scratchy quality that is asso-ciated with Internet telephony today. Callswill be mainlined into the jugular of Qwest’sunlimited bandwidth. Consumers will useQwest’s IP long distance service by making alocal call on a normal telephone, thus access-ing a circuit-to-IP platform by ViennaSystems, a Newbridge Networks affiliate.The Vienna platform will not compress thevoice; it will simply packetize the raw, 64kbps signal and send it via IP. Shafei claimsthat the quality will be virtually as good ascircuit-switched voice.

Several other entrepreneurial companieshave tested the waters for Stupid Network

innovations. Vocaltec, for instance, was thefirst company to commercialize the fact thatyou could use the Internet for voice commu-nication. If you and I both have Internetaccess and we’re running Vocaltec software,we can talk to each other for as long as welike, for no incremental cost, no matterwhere in the world we are. Though voicequality is still not ideal and delay can inter-fere with the flow of a conversation, look forthis new communications niche to mergewith other Internet applications to createnew value.

Placeware, a Xerox PARC spin-off forinteractive multimedia meetings and confer-ences over the Internet, mixes Internet tele-phony with data sharing, presentationgraphics, and a crude representation of the

meeting space. Demos of the technology seemto add a lot to voice conferencing, and it givesa more participatory, less self-conscious feelthan a video conference.

BEYOND QOS TO SIMPLE STUPIDITYIntelligent Networkadvocates point out thatnetworks need to treatdifferent data types dif-

ferently. Right now, they’re absolutely cor-rect. There is a network for telephony,another network for TV, and proprietaryleased-line networks for financial transac-tions — and none of these are ideal for pub-lic Internet traffic. You need to have lowdelay for voice telephony, the ability to han-dle megabit data streams with ease for TV,and low error rates and strong security forfinancial transactions.

Quality of Service (QOS) is an intermedi-ate step in the journey from separate net-works to a single, simple Stupid Network.QOS, in standard telco thinking, means arepertoire of different ways of handling eachtype of data on a single network. If the StupidNetwork is to become a bona fide integratedservice network, it will need to carry all kindsof data with different needs.

But suppose technology improves somuch that the worst QOS is perfectly fine forall kinds of traffic, without a repertoire ofdifferent data handling techniques. Suppose,for example, that everyday normal latencybecomes low enough to support voice tele-phony, while at the same time allowingenough capacity for video, plus data integri-ty strong enough for financial transactions.

MIXED MEDIA FEBRUARY/MARCH 1998n W

Intelligent Network advocates point out that networks must

treat different data types differently. They’re absolutely correct

— for now.

31

This would be a true Stupid Network — onetreatment for all kinds of traffic.

Skeptics might say that there would have tobe dramatic improvements in networking tech-nology for this to happen. Well, we’re gettingthere. Routing switches from Madge andFoundry recently showed performance impres-sive enough to conclude that routing latencyand jitter (variation in packet arrival time) maysoon be a negligible issue. But these were labtests, not field usage, and packet losses were ashigh as 1% under some conditions. So we arenot there yet – but perhaps we will be soon.

PLAYERS IN THE NEW ORDER Still other technologies ofabundance, with thepotential to break the tel-cos’ foot-dragging hege-

mony, have attracted interest fromentrepreneurial vendors. Here are some ofthe leading candidates:✦ LMDS: This technology provides a wire-

less broadband last-mile path to the StupidNetwork. The FCC LMDS auction, com-pleted on March 25, opens U.S. markets todeployment and service over the next cou-ple of years. The two big bidders wereNextband and WNP Communications.Equipment manufacturers Hewlett-Packard, Stanford Telecom, TexasInstruments, Tadiran and others also willbe beneficiaries.

✦ CDMA: Another wireless data accessmethod; Qualcomm is still well positioned.Also watch Broadband CDMA, an emerg-

ing open standard that can deliver fromfractional T-1 on up. Interdigital is the lead-ing B-CDMA player.

✦ Gigabit Ethernet switching: This technol-ogy has moved from the laboratory to themarketplace in a remarkably short time.While Ethernet has been synonymouswith local area networks in the past,“Neighborhood Networks” are replacingATM as the vehicle of choice for campusnets. Can real neighborhoods be farbehind? Gigabit Ethernet players includeBay Networks, Cisco, 3Com, Cabletron,Foundry, Extreme Networks, and manyothers.

✦ Cable modems: Players include set-topbox makers General Instruments andScientific-Atlanta, plus Motorola, HybridNetworks and others. Cable providerComcast is a good bet for cable modemservice — Bill Gates thinks so, anyway (tothe tune of $1 billion).

✦ The power companies: Also worth watch-ing, following Nortel’s announcement lastNovember of technology to deliver data toend users over power lines.All of these infrastructure improvements

are rapidly making the telcos’ IntelligentNetwork a distinctly second-rate choice. Thebottom line, though, is not the infrastructure;it is the innovation that the Stupid Networkunleashes. The Stupid Network ensures thenext paradigm-breaking, market-making“new thing.” The only question is who willbecome the next Netscape, the next Microsoft— or the next Ma Bell. And that’s not a stu-pid question.

MIXED MEDIA FEBRUARY/MARCH 1998n W

Mixed Medi a

PERMISSION TO MAKE DIGITAL OR

HARD COPIES OF ALL OR PART OF THIS

WORK FOR PERSONAL OR CLASSROOM

USE IS GRANTED WITHOUT FEE

PROVIDED THAT COPIES ARE NOT

MADE OR DISTRIBUTED FOR PROFIT OR

COMMERCIAL ADVANTAGE AND THAT

COPIES BEAR THIS NOTICE AND THE

FULL CITATION ON THE FIRST PAGE. TO

COPY OTHERWISE, TO REPUBLISH, TO

POST ON SERVERS OR TO REDISTRIBUTE

TO LISTS, REQUIRES PRIOR SPECIFIC

PERMISSION AND/OR A FEE.

© ACM1072-5220/98/0200

$5.00

~In the next issue of netWorker (April/May), Dado Vrsalovic of AT&T will

present a rebuttal to David Isenberg’s article. We also welcome your

comments on the Stupid/Intelligent Network situation; please send them to

editor@networker.org.