Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until...

12
Ž . Computer Networks 31 1999 157–168 Telephony in the year 2005 Joseph Rinde ) MCI, 2100 Reston Pkwy, Reston, VA 20191-1218, USA Abstract The growth of packet-based voice services is leading to integration of voice and other services over packet switched data networks. This paper explores a possible path that the telephone service industry may follow as this integration is accelerated by technological advances that improve the capabilities of packet-based services while reducing their costs. q 1999 Elsevier Science B.V. All rights reserved. Keywords: IP telephony; Internet 1. Introduction We live in a time of rapid technological change, especially in the field of telecommunications. The term Internet years, a take off on dog years, attempts to capture the speed with which we experience the evolution of new products and services. To attempt in this environment to predict the telecommunica- tions scene seven years hence is a daunting undertak- ing. Surely whatever we predict today will be only a vague perception of the reality to come. No matter how extravagant the prediction, it will pale in com- parison with the reality that will be. With this in mind we can take the trends of today and extrapolate them in a skyward direction, secure that if we are wrong it is likely to be on the conservative side. One of the hottest topics in telecommunications today is the use of packet data networks, and the ) Corresponding author. Tel.: q1-703-715-7163; fax: q1-703- 715-7252; e-mail: joe_[email protected] Internet in particular, to transport voice and fax, traffic that classically runs on the circuit switched Ž . Public Switched Telephone Network PSTN . All of this started three years ago when VocalTec intro- duced a PC application that enabled voice communi- cations between multimedia PCs across the Internet. That first example of affordable phone calls over the Internet was a hobbyist tool. Soon a plethora of vendors offered improved approaches to voice com- munications between PCs, and more importantly be- tween PCs and phones, and between phones using Ž the Internet as the long distance transport see Fig. . 1. The rest of this paper projects future develop- ments by looking at the history of telephony from the vantagepoint of the year 2005. 2. Early history The earliest history of Internet telephony was plagued with poor voice quality and substantial de- lays. These drawbacks were overcome by the allure 1389-1286r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. Ž . PII: S0169-7552 98 00266-9

Transcript of Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until...

Page 1: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

Ž .Computer Networks 31 1999 157–168

Telephony in the year 2005

Joseph Rinde )

MCI, 2100 Reston Pkwy, Reston, VA 20191-1218, USA

Abstract

The growth of packet-based voice services is leading to integration of voice and other services over packet switched datanetworks. This paper explores a possible path that the telephone service industry may follow as this integration is acceleratedby technological advances that improve the capabilities of packet-based services while reducing their costs. q 1999 ElsevierScience B.V. All rights reserved.

Keywords: IP telephony; Internet

1. Introduction

We live in a time of rapid technological change,especially in the field of telecommunications. Theterm Internet years, a take off on dog years, attemptsto capture the speed with which we experience theevolution of new products and services. To attemptin this environment to predict the telecommunica-tions scene seven years hence is a daunting undertak-ing. Surely whatever we predict today will be only avague perception of the reality to come. No matterhow extravagant the prediction, it will pale in com-parison with the reality that will be.

With this in mind we can take the trends of todayand extrapolate them in a skyward direction, securethat if we are wrong it is likely to be on theconservative side.

One of the hottest topics in telecommunicationstoday is the use of packet data networks, and the

) Corresponding author. Tel.: q1-703-715-7163; fax: q1-703-715-7252; e-mail: [email protected]

Internet in particular, to transport voice and fax,traffic that classically runs on the circuit switched

Ž .Public Switched Telephone Network PSTN . All ofthis started three years ago when VocalTec intro-duced a PC application that enabled voice communi-cations between multimedia PCs across the Internet.That first example of affordable phone calls over theInternet was a hobbyist tool. Soon a plethora ofvendors offered improved approaches to voice com-munications between PCs, and more importantly be-tween PCs and phones, and between phones using

Žthe Internet as the long distance transport see Fig..1 .

The rest of this paper projects future develop-ments by looking at the history of telephony fromthe vantagepoint of the year 2005.

2. Early history

The earliest history of Internet telephony wasplagued with poor voice quality and substantial de-lays. These drawbacks were overcome by the allure

1389-1286r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0169-7552 98 00266-9

Page 2: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168158

Fig. 1. Communications across the Internet between PCs, PC and phone, and between phones.

of free phone calls oÕer the Internet. Of course thecalls were not free, but they were inexpensive com-pared to PSTN charges. In the case of internationalcalls prices were as low as 25% of discounted PSTNrates. The commercialization of this cost disparityled to an array of services that arbitraged the cost ofInternet telephony against the artificially high pricesof regulated international PSTN services. The service

Žquality could be lower than the PSTN round trip.delays sometimes as high as 3 seconds and still

attract customers willing to trade quality for a lowerprice.

Many questioned the ability of the Internet todeliver voice quality on a par with the PSTN. Delaywas the big issue, for within two years of the incep-tion of this technology, codecs were introduced thatproduced voice quality comparable to the Pulse Code

Ž .Modulation PCM form of digitization used in thePSTN, at a fraction of the standard 64 kbps full

w xduplex transmission used in the PSTN 1 .The first use of voice over IP was made with PCs

at each end. The PCs introduced delay from a num-ber of sources. The primary source was the soundcard. Early sound cards were half-duplex and weredesigned to play sound sources from CD-ROMs.Given the multi-second delay in loading the softwareand getting data off early CD-ROM drives, soundcard designers saw no need to optimize the addi-tional delay they introduced, which was typicallyless than a second. However, delays of a half-secondin the sound card alone equaled the round trip delay

in satellite telephony. Additional delay came fromthe 486 Intel processors in the PCs used in thoseearly years. Not only was the processor compara-tively slow, but the predominant operating systemŽ .Windows 3.1 was not a true multi-tasking system,which introduced delays as it emulated multi-tasking.The early uses were frequently made with dial accessto the Internet. Even using the newest modems of

Ž .that time 28.8 kbps noticeable serialization delaywas introduced at each end of the connection.

In addition to the end point delays the best effortnature of the Internet introduced variable delays andeven gaps in the transmission of voice that resulted

Ž .in a perceptibly lower Quality of Service QoS thanon the PSTN.

Delay was sometimes tolerable in two party con-versations, but in multi-party conference calls delaywas very disruptive as multiple parties vied to speaknext. Delay in this environment resulted in unin-tended interruptions and an awkward back-off pro-cess to resolve who would continue speaking. Evenin two party conversations, delay was more readilyaccepted by price conscious consumers than by busi-ness users.

3. From 1998 to 2001

Although there were some early applications ofInternet telephony in 1996 and 1997, it was not until

Page 3: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168 159

1998 that services based on this new technologybecame significant.

The continued explosive growth of data capacityduring this period took bandwidth used for data fromone tenth that used for the PSTN to full parity. Thisgrowth included the public Internet, and corporate

Žintranets of various technologies IP, Frame Relay,.ATM, etc. . Some of this growth was fueled by the

migration of voice from the PSTN to data networks.This period saw the evolution of carrier grade

Ž .high reliability and scalable services based on IPcarriage of voice. Not only did the delay problem getresolved but also a whole new set of services cameinto being. Carrier grade equipment made it econom-ical to build large systems to accommodate the ex-panding customer base. This equipment also pro-vided the high reliability needed for a mass telecom-munications service.

No longer was IP telephony only a cheaper wayto make calls, but a better way to make them. Theimplementation of the World Trade OrganizationŽ . w xWTO 2 agreement to bring competition into inter-national telecommunications brought Internet tele-phony and PSTN prices to a cost basis, and on anequal regulatory footing. With price parity on theregulatory front IP telephony took advantage of the

Žlower costs of transmission lower switching costs,.and reduced bandwidth usage to offer a competitive

service for normal telephony. IP telephony addition-ally offered services not available on the PSTN, acombination that sped the deployment of IP tele-phony.

3.1. Fax

The PSTN was also used to transmit documentsusing fax machines. These machines turned scanneddocuments into data for transmission over the circuitswitched PSTN. Early offerings in this field enabledthe transmission of fax over the Internet. The earliestofferings did not integrate voice and fax support,rather concentrating on vertical applications of trans-mitting traditional PSTN traffic over IP.

Transmission of fax over the Internet introduced adifferent, though no less demanding set of designchallenges. In the voice arena delay was king, whilepacket loss, if not too high, was tolerable. Lostpackets represented a 30 to 90 millisecond loss of

speech. The terminating Internet equipment, whetherPC or gateway, typically extrapolated from the lastsound received to fill the void. Human intelligence atthe receiving end was usually able to make up forthe missing sounds. In the worst case, too many lostpackets resulted in the informal retransmission proto-col ‘‘what did you say?’’

Fax on the other hand could tolerate a moderateŽamount of delay the T.30 protocol used by fax

machines specified how much delay was acceptable.before the call failed , but it could not tolerate lost

bits. Retransmitting lost packets across the Internetcould insure no loss of data while various tricksfrom the T.30 protocol could be used to stall for timewhile the retransmission was taking place. Othercomplications of fax over the Internet involved mo-dem handshaking. Some of these functions weredone from the near end gateway for efficiency andfor timing necessities, while other functions wereallowed to happen between the end fax machinemodems.

3.2. So what was better about IP telephony?

The earliest useful IP telephony was phone tophone service. Although the number of PC userscontinued to grow they still represented a smallpercentage of the population, compared to the num-ber of telephone handset users. With phone to phoneservices customers continued to use their existingtelephone handsets, calling local access points toenter the IP telephony system. After dialing the localaccess number the caller entered his billing numberfollowed by the destination telephone number. Thiswas reminiscent of the multi-stage dialing of thecompetitive long distance carrier offerings beforeequal access.

In 1984 the United States required the local tele-phone companies to offer equal access, allowingcustomers to pre-select which carrier would carrytheir long distance calls. This allowed customers ofall long distance carriers to dial calls the same way,without the need for additional digits. This made itpossible for carriers other than AT&T to offerequally convenient services. Before equal accesscompetitors were restricted to offering multi-stagedialing which limited their ability to penetrate the

Page 4: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168160

market. As an example when equal access started in1984 MCI, after 20 years of effort, had only 4% ofthe US domestic long distance market. By 1998,after 14 years of equal access, MCI had 20% of themarket.

The delay problem was solved in several ways.For phone to phone service gateways with dedicatedDSP chips minimized the processing delay. High-speed connections to backbone transport networksminimized the transmission delay. Some carriers ac-complished this by building separate networks forvoice traffic, carefully engineering those networks tokeep queuing delays to a minimum. Other carriersimplemented priority queues to allow a mixture ofvoice and other traffic, less demanding of real timeperformance. The latter allowed economies of scaleto be applied on the aggregated traffic. Careful engi-neering of the network was required by both ap-proaches.

Two other factors contributed to the cost-effectiveimplementation of IP telephony in this time period.First, the cost of IP equipment came down. Ever-big-ger routers at ever lower per bit costs drove the,already low, cost per switched bit down by twoorders of magnitude at the beginning of this timeperiod and another two orders of magnitude by theend of this time period. Similarly, the gateways usedto interface PSTN initiated calls to the IP transportcame down in price by one order of magnitude. Thiscost was further reduced by IP enabled handsets, thatincorporated the gateway function in the customerequipment, removing that factor entirely from thecarrier’s cost equation when servicing such cus-tomers.

The second key factor was a reduction in band-width cost on international circuits. With the intro-duction of new optical cables with far higher capaci-ties, and the retrofitting of some existing opticalcables to increase capacity, the abundance oftransoceanic bandwidth brought prices down. Thismade it feasible to cost effectively engineertransoceanic IP networks to provide the same levelof low delay voice carriage that could be providedon continental service.

The combination of faster processors and lowerpriced bandwidth enabled IP telephony in this periodto deliver CD quality sound between IP phones.

Soon customers started complaining about the poorquality of voice over the PSTN.

3.3. The year 1998

At the beginning of this period true costs of voicecarriage over IP and PSTN networks were very

w xsimilar 3 . Where IP carriage used less bandwidthover continental routes and IP switching was a little

Ž .cheaper by less than a factor of 2 the high cost ofthe gateways almost eliminated the cost advantage ofIP carriage. Arbitrage was the operative driver. Us-ing Internet transport regulatory costs, such as accesscharges, were circumvented. In the United States thisreduced the comparative cost of IP carriage by anadditional 4 cents a minute. This was a huge advan-tage given that the cost of transport was less than 1.5cents a minute. The savings on transport were almostinsignificant compared to the savings on accesscharge avoidance. Other costs, such as sales andmarketing, network operations, and customer serviceall overshadowed the transport cost as elements inthe cost equation.

Local pickup and delivery of calls used up part ofthe savings from access charge avoidance. The IPlong distance carrier was forced to use retail ser-vices, in the form of business lines, to provide localaccess. Since the IP telephony provider was receiv-ing calls at the access gateway there was no per-

Žminute charge for the call. In the PSTN paradigmthe caller pays for the call. Since the origin gatewayreceiÕes calls it has no per minute charges associated

.with the incoming calls. The service provider’s perminute cost was the monthly cost of the business linedivided by the number of minutes of actual usage.The cost of the gateway port was similarly convertedto a per-minute cost by dividing the monthly depre-ciation by the actual number of minutes used.

ŽThe technology of routers remained one step al-.beit a small step ahead of the demands of Internet

growth. The period started with ATM based subnet-Ž .works operating at OC12 622 Mbps speeds inter-

Ž .connecting routers operating at OC3 155 Mbpsspeeds providing the IP backbone. It was in 1998that IP over SONET became a reality. Until this timerouters were limited to OC3 and lower speeds. Al-though OC3 speeds were considered substantial inthose days, the number of OC3 ports on router was

Page 5: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168 161

limited to one or two, not enough to build a networkwith. During 1998 and 1999 the routers began run-ning directly over the SONET network, first at OC12

Ž .speeds and then at OC48 2.4 Gbps . This not onlyallowed the backbone to keep up with capacity de-mand, but also reduced overhead by removing theATM cell based subnetwork.

In the year 2000 IP over fiber bypassed theSONET network, using IP’s routing flexibility inplace of the SONET dual fiber rings for reliability.This freed up otherwise unusable bandwidth in thefiber network.

3.3.1. DelayThere remained the problem of delay. In 1997

many thought that reservation protocols, such asRSVP would solve the delay problem. This provedto be partly true. Reservation protocols required thenormally stateless routers to recognize and monitordata flows. Although this worked well in small-scalenetworks it was impractical in the Internet backbone,where routers handled thousands of flows per sec-ond.

RSVP was applied successfully in corporate net-works, as will be discussed further below. RSVP wasalso effective in access networks, such as those

Ž .operated by Internet Service Providers ISPs . Theproblem, as noted above, was support of RSVP inthe backbone.

Differentiated QoS in the backbone was solvedusing priority queuing and various approaches totreating the queues differently. Placing voice packetsin the high priority queue allowed those packets tomove ahead of web and other file oriented traffic.Special treatment of the high priority queue in con-gested conditions allowed other packets to be lostbefore voice or fax traffic began to suffer. Thismanifested itself in improved voice quality as thereceiver had fewer occasions to invoke its compensa-

Žtion algorithms to deal with lost or significantly.delayed packets.

At the ISP to backbone interface RSVP character-istics were converted to prioritized traffic for trans-port across the backbone. The reverse conversionwas performed at the interface to the destination ISPfor delivery to the receiving IP telephony software.In this way low delay and low packet loss wasreasonably assured, but at a price.

The attraction of lower cost transport using theInternet was impacted by the cost of providing ahigher QoS transport. RSVP was not free to the ISPswho had to provision additional bandwidth to allowRSVP based services without serious impact to theperformance of other data flows. This higher costwas passed to the RSVP requester and incorporatedin the price of Internet telephony service. Similarlythere was a cost for the use of prioritized traffic inthe backbone. This service also required the provi-sion of additional bandwidth to provide acceptableservice to non-prioritized traffic. This cost was alsopassed on to the priority service requestor and addedagain to the cost of providing higher quality IPtelephony.

By late 1998 two tiers of IP telephony appeared inthe public marketplace. A high QoS version based onpriority transmission, and lower QoS for cost con-scious consumers. Additional service providers tar-geted both the consumer and the business marketsthat year.

The services were all based on a two-stage dialusing a local access number. This requirement to diala local number, followed by a billing number, fol-lowed by the destination telephone number limitedthe market penetration of the services, as most callersdid not want to be bothered with the extra dialing. APBX could provide such services transparently to thecaller by automatic call diversion in a PBX. ThePBX could route the call, based its least cost routingtables to a process that prefixed the dialed digits withthe local access number and the billing number.Despite this ability few telecom managers were will-ing to do this. The reasons for this resistance werevaried, but generally were related to quality concernsof mission critical communications over new ser-vices that were not under the telecom manager’scontrol. Sometimes this was a political consideration,especially when different departments were responsi-ble for the data and voice services. Other times itwas a purely practical consideration based on limitedmanagement tools for tracking and trouble shootingvoice problems on the data network.

3.3.2. Intracorporate communications1998 saw the first large wave of usage of IP

telephony for internal corporate communications.Some of this deployment was aided by PBX vendors

Page 6: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168162

adding IP telephony capability to their product offer-ing, but more often it was the independent additionof an IP telephony gateway between the PBX andthe data LAN.

ŽFor well-engineered enterprise networks meaningWAN interconnection of LANs using IP over leasedlines, Frame Relay, or ATM with WAN line utiliza-

.tion below 50% IP telephony was added at a lowmarginal cost. At less than 20 kbps, half-duplex, pervoice session the impact on LAN usage was negligi-ble, especially as LANs moved to gigabit speeds.The same applied on lightly loaded WAN connec-tions in excess of 500 kbps. This allowed inter-officecalling using the enterprise data network as the tie

Ž .line between the PBXs Fig. 2 . The use of inter-office IP telephony was transparent as the PBX wasused to automatically route calls over the enterprisedata network. Based on the incremental cost of usingthe enterprise data network per minute costs forinter-office calls dropped below 5 cents a minute,which was lower than otherwise available PSTNservices.

The cost savings came with the pain of earlyadoption. Bugs and performance issues attended theinstallation of intranet based IP telephony. Some

companies solved the performance problems by in-troducing RSVP, while others applied directed net-work engineering to allow the best effort nature ofthe routers to deliver acceptable service levels. Thebiggest unfulfilled promise of corporate based IPtelephony was in the provision of international tele-phone service. The high cost of international data

Žtransport whether leased lines or data services such.as Frame Relay made engineering of the network an

expensive proposition. Where international links inthe intranet were too heavily loaded, the delaysintroduced limited the usefulness of IP telephony inthis area.

1998 also saw the first uses of the so-called softPBX. This was the first generation of the IP basedPBX. Some of these products supported commonwire line phones while some supported IP enabledphones. It was the latter that brought about the

Ž .greatest changes Fig. 3 .Soft PBXs could replace more expensive propri-

etary PBXs with equipment based on industrial PCplatforms. The interface of the soft PBX to commonwire line phones was done through Computer Tele-

Ž .phony Integration CTI hardware that had beendeveloped for call centers and other Interactive Voice

Fig. 2. Use of IP gateways to divert PBX traffic over the enterprise network.

Page 7: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168 163

Fig. 3. Soft PBX with IP phones.

Ž .Response IVR applications. The wiring to thephones remained the same, but inexpensive phonescould be used in place of pricey proprietary unitsfrom the traditional PBX manufacturers.

This solution suffered from the loss of extrafunction buttons and lights that were available on theproprietary phones. The attempt to compensate for

Žthis with PC support on the assumption that every.worker had a PC as well as a phone on his desk

failed to catch on. Although the PC adjunct to thephone service could provide more functionality thaneven the proprietary phones from classic PBXs, itrequired a significant change in user behavior, whichslowed adoption.

The truly significant impact of the soft PBX in1998 was the introduction of the IP enabled phone.Although the penetration of these devices was verysmall that year, it set the stage for a massive changeto come in later years. No longer was it necessary tochange user behavior when switching to a soft PBX.

The IP phone connected to the LAN, bypassingtelephone wiring. This removed the CTI hardwarefrom the soft PBX, reducing its price to the cost ofthe PC plus the software license. The gateway neededto connect to the PSTN was either incorporated inthe soft PBX or provided by another box. The indi-vidual IP phones communicated with the soft PBX to

place calls to other IP phones, or through the IP toPSTN gateway to any phone worldwide. Thesephones started in the US$ 300 to US$ 400 pricerange, which was considerably less than similarlyfeatured phones for conventional PBXs.

The soft PBX included the usual PBX features ofthat time plus it provided APIs enabling new featurecreation.

3.4. 1999 to 2001

These years saw significant maturation of the IPtelephony technology. They also saw the relativestabilization of the market players. The traditionaltelecommunications companies and telecommunica-tions equipment suppliers that adapted came to thefore along with the well-managed startups in the IPtelephony industry. These events were accompaniedby the arrival of lower cost transoceanic fiber capac-ity.

The solutions to the delay problems that wereintroduced in 1998 became widely deployed, andrefined. By the end of this period routers were

Ž .running OC192 10 Gbps directly over fiber, en-abling the Internet backbone to support multimediacommunications with acceptable delays. The newfiber capacity that became operational in 1998 started

Page 8: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168164

to make an impact on the prices of transoceanicbandwidth. This enabled the economic provisioningof sufficient bandwidth to support high QoS voiceservices on transoceanic IP networks.

3.4.1. Promises promisesFrom its earliest days the real promise of IP

telephony was not cheaper voice but a combinationof an integrated network and new compelling appli-cations that would bring users flocking to the tech-nology. It was in this time period that these promisesstarted to be fulfilled in a meaningful way.

The IP enabled phone came into its own in largepart as a result of new services it offered that couldnot be matched by conventional phones. The ground-work of infrastructure had been laid in 1998 as

Ž . w xvarious forms of Digital Subscriber Line xDSL 4and cable modem services were deployed to bothconsumers and businesses. Once high bandwidth ser-vices, at affordable prices, were available over exist-ing wires to the premise, the entrepreneurs were ableto work their magic. All of this was true in spadesfor commercial enterprises that had been wired withfiber.

w xFor residential service ADSL 5 solved the issueŽof lifeline service phone operation in emergencies

.such as power failures by using existing phonetechnologies. For everyday use the high speed ofADSL provided a vehicle for IP telephony and a hostof related services. ADSL allowed the use of Plain

Ž .Old Telephone Service POTS over the same wirepair providing lifeline service. Where cable modemswere used, a separate POTS line was also needed toprovide lifeline service. This put cable services at adisadvantage in areas that offered ADSL services.

3.4.2. Metamorphosis in the central officeAs the traffic migrated from circuit switching to

packet switching the equipment in the Central OfficeŽ .CO became heavily packet oriented. Large routerswith gigabit transmission made delivery of low la-tency data to the end user a reality. The CO became

Ž .the hub of ISP activity Fig. 4 .The nature of xDSL made it possible to provide

competitive services. Any ISP could connect with ahigh-speed pipe, through the CO, into the userpremise and offer differentiated service. The largerISPs from the 90s entered this arena along with thecompetitive local telephone companies. The latterwere predominantly large telecom companies, theclassic long distance carriers and local telephonecompanies moving into other areas.

3.4.3. IP enabled phonesThis was the era of the IP enabled phone. With

increased volume the prices of these soft PBXspawned devices came down to the consumer levelŽ .under US$ 100 . Common wire line phones re-mained cheaper but could not compete in the func-tionality play. With the availability of high band-

Fig. 4. The central office of the early 2000s.

Page 9: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168 165

width and the ever-dropping cost of computing, thesedevices became capable of delivering not only lowdelay voice, but also higher quality voice than thePSTN. Codecs to take advantage of the availablebandwidth and processing power delivered CD qual-ity voice and music. This raised the public’s expecta-tions of sound quality in telecommunications, mak-ing the conventional PSTN appear old.

Integration with universal mailboxes allowedhome users to enjoy the same services PBX usershad with message waiting lights and preprogrammed

Ž .feature buttons Fig. 5 . Both home and office usersenjoyed the enhanced capabilities of integrated mail-

Ž .boxes of voice, fax and data e.g. e-mail with theirhand set. Although a PC or high end IP enabledphone was required to take advantage of the fullcapabilities of these mailboxes, most users weresatisfied with simpler versions of the capabilitiesmade available by the less expensive versions ofthese phones.

xDSL services brought with them LANs. Thisintroduced LANs into many homes and also intosmaller offices. These inexpensive, older technology,100 Mbps LANs, although slow by comparison tothe gigabit LANs in more sophisticated offices,opened up new options for both home and office.With the cost of the LAN included in the xDSLservice additional IP enabled appliances could beinstalled at a low incremental cost. In the consumerspace this appeared in the form of alarm systemcommunications, health alarms for the elderly and

the seriously ill, as well for the convenience-gadgetoriented consumer. This facilitated the introductionof devices to monitor your weight, your consumption

Ž .of vitamins to automatically reorder , and manyother every day tasks that we take for granted today.

Small offices that previously had no data infras-tructure suddenly found new functionality open tothem. Where limited key systems were used before;the IP enabled phone permitted Centrex and PBXlike features previously only available at consider-able expense. Just as the World Wide Web helpedlevel the playing field for small companies to createa worldwide presence, so too IP enabled phonescoupled with competitive service providers gavesmall companies the same productivity tools enjoyedby the largest of organizations.

3.4.4. How usage changedAlthough video conferencing made the work place

more productive it still failed to have high penetra-tion in the residential market. Social forces kept theintrusive technology at bay in the home market.

The integrated functionality to mix voice, videoand shared data with the touch of a button had ahuge impact on business communications. In oldertimes one could drop in on a colleague down the hallto collaborate, but the same did not apply to col-leagues at other locations. Limited to voice commu-

Žnications and out of band data sharing using fax,.e-mail or web page the value of collaboration across

the miles was less than down the hall. With the

Fig. 5. Home configuration of the early 2000s.

Page 10: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168166

availability of easily setup multimedia calls thischanged. Now workers could knock on a colleague’svirtual door across the globe just as easily as knock-

Ž .ing on a real door or cubicle wall down the hall.The ensuing conversation could be carried out withshared data and visual feedback. Even more impor-tant was the ability to easily add other parties to themultimedia call. This duplicated the natural humanbehavior of adding others to impromptu meetings ina building when additional colleagues were neededto resolve an issue. Once these capabilities could beinvoked by anyone with a PC the face of communi-cations changed, for communications both inside andoutside the company.

Connections to some parts of the world did notprovide adequate voice quality, primarily due todelays. To support these locations conferencing sys-tems included the ability to use the pre-existingPSTN capabilities to provide the voice part of themultimedia call. This use of the PSTN was donetransparently by the caller’s PC requesting PSTNservices once the IP connection proved unable todelivery the required QoS.

3.4.5. What opened the damPerhaps the biggest impact on the economics of

IP telephony was the drop in price of the IP enabledphones. The lower price made these devices afford-able by large numbers of consumers. The most dra-matic effect of this development was the reduction inthe growth of gateway ports needed by IP telephonyproviders. Until this time calls were predominantlyinitiated by analogue handsets connected to the legacyPSTN. This required relatively expensive equipmentto convert the PSTN style call to an IP telephonyformat for transmission across the IP network. At theother end another gateway was needed to reconvertthe call for delivery to a PSTN connected handset.

Calls initiated or terminated by IP enabled phonesdid not require a gateway to connect the call to thatdevice. Gateways were still needed to address thelarge installed base of analogue handsets. Indeed thisinstalled base continues to dominate the market tothis day. Nonetheless the spread of IP enabled phonesslowed the growth of gateways in the IP telephonysystems, providing a beneficial economic effect forservice providers and customers.

Partly fueled by the growth of IP telephony thetotal bandwidth used for data networks grew equal tothe bandwidth used by the PSTN worldwide. This setthe stage for the next steps in the evolution of thetelecommunications industry.

3.5. The years 2002 to 2005

By 2002 IP telephony systems’ success in thedeveloped countries led to rapid adoption of thismature technology in the developing world. Adopt-ing countries heard their scratchy telephony servicestransformed into crystal clear high quality sound.Running fiber throughout urban areas brought thesebenefits to the city dweller. The rural populationcontinued to build phone service using cellular tech-nologies. Here too positive consequences from thematuration of IP telephony technology were found.These same technological developments producedhigher quality sound at lower costs than were avail-able in cellular service of the late 1990s.

The relentless march of Moore’s law continued todrive down the price of IP communications, both forIP enabled phones and for PCs. Together these lowerprices extended the reach of collaborative multime-dia. The result was a paradox of social impact. Onthe one hand it brought diverse people together andon the other hand it accentuated the identity ofgroups.

Communities of people, whether of a nationalist,religious or racial nature, were strengthened acrossgeographic boundaries. Not only were geographicdistances removed as obstacles, but also politicaljurisdictions became less of a barrier. Inexpensiveand intimate communications allowed extendedgroupings to develop stronger bonds. It was thecombination of lower cost of bandwidth coupledwith more efficient usage of that resource, plus theincreased computing power that allowed the deliveryof affordable technology for intimate communica-

Žtions. It was video and data sharing in the form of.photographs, videos and live events that drove this

trend. The lack of control of incoming video callsheld back the video phone vision of the 1950s in theform of a phone with a picture in place of the blackphone of yesteryear. Rather the videophone of theyear 2003 was more like the camcorder of the 1990s,

Page 11: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168 167

Žbut in real time allowing selective use under the.control of the party being viewed . Thus many homes

had one and some had many.

When using videophones, each party had the abilityto turn off the video transmission at its end of thecall, enabling the use of video only when required orwhen a close friend called. This was a sociallyawkward function, conveying strong signals aboutwho the controlling party considered a close friendand conveyed even stronger signals when the statusof closeness changed for the worse. This createdsocial stresses that slowed the spread of this technol-ogy.

In a seeming paradox cheaper multimedia com-munications brought people of different groups to-gether through the same intimate communications.Although this was primarily in the business environ-ment, it spilled over into the social realm as officeparties brought workers together with their familiesacross geographically dispersed locations.

The continued drop of xDSL prices coupled withincreased speeds expanded the types of services of-fered to both consumers and small businesses.

Inexpensive Internet appliances made the smarthouse affordable to millions of consumers across theworld. Using an IP enabled phone a consumer couldjust as easily adjust his alarm system remotely asstart his bathtub filling at a prescribed temperature.Bar code readers in refrigerators and pantries helpedconsumers keep track of inventory and locally flagthe consumer to the need to reorder. For busy peoplethe IP appliance could automatically reorder items inshort supply from a nearby store.

The increased use of IP transmission resulted infaster conversion of telephone central offices to IPbased services. By the end of this period a quarter ofall telephone services used packet technologies. Im-proved battery technologies introduced in 2003 made

Ž .uninterruptable power supplies UPS inexpensiveand long lasting. This made some consumers com-fortable enough to eliminate the POTS connectionthat was becoming a relatively larger part of the totalphone bill. This reduced the pressure on circuitswitched growth in the CO. This is a trend thatshows no sign of abating.

It was also in this time period that the expansionof the worldwide telephone network was predomi-

nantly done by expanding the IP portion of the longdistance network. The increased use of IP in the COlowered switching costs, and improved bandwidthutilization. It also addressed the growing demand forend to end IP connectivity to facilitate the use of

Ž .new functionality such as better quality voice acrossthe network.

4. Conclusions

Having peered into the future of telephony we seean acceleration of the merging of circuit switched

Ž .traffic voice and fax with packet switching tech-nologies. Integrated services, once envisioned in the

Ž .circuit switched world remember ISDN? will cometo fruition in the packet switched world. It may bethat the IP telephony of today will become theunplanned source of the ultimate integration of voiceand data.

The proponents of packet switching should notbeam too brightly, for the flexibility of the packetworld is a two edged sword. Although anything ispossible, given the right software and enough pro-cessing cycles, there remains the demand of back-ward compatibility. People want to be able to pickup any phone type device and call any other phonewithout worrying about what kind of hardware ofsoftware is at either end or in the middle. This is thechallenge that the industry must meet to make areality of the future described here.

References

w x1 R.V. Cox, Three new speech codecs from the ITU cover aŽrange of applications, IEEE Communications Magazine Sep-

.tember 1997 .w x2 http:rrwww.wto.orgrpressrsummary.htm.w x3 J. Rinde, The future direction of Internet telephony and its

effect on business, Fall VON 97, Boston, http:rrpulver.comrvon97rfall97rrequest.htm.

w x Ž .4 Committee T1, High-bit-rate digital subscriber line HDSL ,TR No. 28, February 1994; also Generic requirements forhigh-bit-rate digital subscriber lines, Bellcore TA-NTW-001210.

w x5 ANSI T1.413-1995, Network and Customer Installation Inter-Ž .faces – Asymmetric Digital Subscriber Line ADSL Metallic

Interface.

Page 12: Telephony in the year 2005hgs/papers/2072.pdfInternet telephony in 1996 and 1997, it was not until J. RinderComputer Networks 31 1999 157–168() 159 1998 that services based on this

( )J. RinderComputer Networks 31 1999 157–168168

Joseph Rinde holds a Bachelor of Sci-ence with High Honor from Stevens In-stitute of Technology and a Master ofScience in Computer Science fromCarnegie Mellon University. From 1974until 1983 Joe was architect of the Tym-net network as it grew from 150 nodesto over 1500 nodes. He wrote the Super-visor, the central control program for thenetwork and later managed the develop-ment of all the network internals as wellas all the CCITT interfaces. From 1983

to 1985 Joe was Director of Advanced Product Development atAmdahl’s data division. From 1985 to 1989 he was a marketingDirector at Equatorial Communications, the pioneer of the VSATindustry. From 1990 to 1991 Joe was Vice President of marketingfor PEER Networks, a startup that designed a network basedcomputing platform. From 1991 to 1993 Joe was an independentconsultant. Among his assignments was a key roll in the develop-ment of MCI’s Hyperstream Frame Relay network; a study ofsoftware quality assurance procedures by major US softwaredevelopment companies; development of standards for collectionand exchange of accounting information for cellular systems; anda design for the integration of Tymnet and MCI’s IP network. In1993 Joe joined MCI full time as a Director in the data servicesengineering department. In 1995 he joined Vint Cerf’s data archi-tecture group where he has worked on universal dial in networks,world wide web based services and IP phone and Fax services.