The Evolution of Internet Telephony
As a founding member of the new Internet Operators Group, AT&T holds a vast amount of experience and expertise in building and operating IP-based networks. We participate in North American Network Operators Group (NANOG), the Internet Engineering Task Force (IETF) and Internet2 in addition to sponsoring the AT&T Center for Internet Research and the AT&T Interoperability Lab. We intend to be the number one global provider of communications services delivered over the Internet by pursuing a worldwide, facilities-based strategy.
Recently, AT&T has taken aggressive steps toward realizing this vision. Teleport Communications Group (TCG) brings to AT&T......
White Paper:
The Future of Online Communication
Section I: The Evolution of Internet Telephony
H. Mine, Executive Vice President, Probe Research, Inc.
Section II: AT&T’s IP Telephony Initiatives
October 1998
Section I: The Evolution of Internet Telephony
Hilary Mine
Probe Research, Inc.
Three Wing Drive, Suite 240
Cedar Knolls, NJ 07927
www.proberesearch.com
Tel: 973-285-1500
September 1998
TABLE OF CONTENTS
VOICE AND DATA – CONVERGENCE HAS BEEN A LONG TIME COMING ...........................................4
EMERGENCE OF THE INTERNET........................................................................................................................6
INTERNET TELEPHONY - MARKET EVOLUTION AND DRIVERS............................................................8
HOBBYIST PHASE - IP TELEPHONY AS HAM RADIO SUBSTITUTE .....................................................................................9
TARIFF ARBITRAGE - CALLING CARD REPLACEMENT....................................................................................................9
TARIFF ARBITRAGE - ENTERPRISE TRUNK REPLACEMENT.............................................................................................11
FEATURE PARITY.................................................................................................................................................11
NEW APPLICATIONS..............................................................................................................................................12
Conferencing Applications.................................................................................................................................12
Call Center Applications....................................................................................................................................13
IP Messaging......................................................................................................................................................14
IP Call Waiting...................................................................................................................................................14
COST PARITY......................................................................................................................................................14
MARKET AND TECHNOLOGY BARRIERS TO INTERNET TELEPHONY AND THE NEW PUBLIC
NETWORK.................................................................................................................................................................15
INTEROPERABILITY CHALLENGES.............................................................................................................................15
QUALITY OF SERVICE............................................................................................................................................16
COST TO DEPLOY IPT..........................................................................................................................................17
OPERATIONAL SUPPORT SYSTEMS............................................................................................................................17
REGULATION - THE VERY REAL THREAT.....................................................................................................17
SERVICE MARKET OPPORTUNITIES IN INTERNET TELEPHONY.........................................................19
THE VALUE OF EARLY MARKET LEADERSHIP/ TODAY’S INTERNET TELEPHONY SERVICE
PROVIDERS...............................................................................................................................................................20
AT&T IP TELEPHONY INITIATIVES.................................................................................................................23
CONSUMERS........................................................................................................................................................23
CARRIERS...........................................................................................................................................................23
BUSINESS CUSTOMERS..........................................................................................................................................24
RESEARCH INITIATIVES..........................................................................................................................................24
AT&T ADVANTAGE............................................................................................................................................24
SUMMARY..........................................................................................................................................................25
Voice and Data – Convergence Has Been A Long Time Coming
The promise of a single network carrying both voice and data is as old the first data networks
themselves. The fundamental notion that a single network offers economic advantages, both in
terms of upfront equipment costs, and with respect to management and administration costs,
goes back literally decades. Perhaps the most famous failure of convergence was ISDN
(Integrated Services Digital Network, or “It Still Does Nothing” as it is colloquially referred to),
which is finally beginning to be deployed today, but which has been available to service
providers since the mid 1980s.
Just as its name implies, the idea behind ISDN was integration of voice and data over the public
network. The expectation was that a customer would use one 64 Kbps B channel for voice, and
the other for data. But carriers did not deploy the technology in any significant way in the U.S.,
largely because their business structures prevented them from being able to prove in business
cases. In addition, as is generally true of new technologies, the developments of true
interoperability and standards, as well as the cost of early customer premise equipment (e.g.,
modems) were also significant barriers to the success of ISDN.
Finally, however, there was a lack of mass market demand for data networking. This particular
market barrier remains, of course, subject to much debate. Despite the fact that 45% of US
homes have at least one PC in use, and the majority of those are Internet subscribers, there are
still those in traditional carriers (e.g. BellSouth) who remain convinced that end users are not
willing to pay the true costs to provide data network services, with the exception of a few large
business customers.
With the advent of the Internet, there is little doubt that end users in both the consumer and
business markets are very interested in data services. Willingness to pay has still to be proven in,
since the mass market Internet is arguably being subsidized at this time. Nonetheless, data
traffic even prior to the rise of the Internet, has been growing dramatically faster than voice
traffic, and this pattern is at the crux of the integrated network dream.
Figure 1: GLOBAL NETWORK TRAFFIC
Terabytes/Yr
1997 1998 1999 2000 2001 2002 2003 2004 2005
PSTN VOICE PACKET & PRIVATE WAN VOICE
PSTN DATA PACKET & PRIVATE WAN DATA
Source: Probe Research, Inc. estimates
Figure 1 shows Probe's estimates of total global communications traffic. PSTN refers to the
public switched telephone network, or the traditional telephone network. Private Wide Area
Networks (WANs) refer to leased line and virtual private networks (VPNs) that large companies
and institutions use for internal communications. Packet refers to packet network based traffic.
Today, over 70-75% of this traffic is still voice, but as we have observed, data is growing much
faster. By the year 2005, data will account for nearly half of global traffic. It is critically
important to note, however, that voice traffic will be growing as well over this period, and will
still be the bulk of traffic on many routes.
A few service providers have begun to observe portions of their networks reaching the magic
50% data traffic threshold, and anticipate that on some segments (particularly international
routes), data traffic will account for as much as 80% of traffic within the next 5-7 years. This is
generally true of routes (e.g., Tokyo to London) that are dominated by business traffic. As a
rule, the more local the traffic, the more it is dominated by residential and small business traffic,
and consequently, the more it is dominated by voice versus data traffic.
In the long run, multimedia applications that are targeted to consumers, such as interactive
video, Internet TV, videoconferencing, and more, will radically alter the mix of voice and data.
Once the standards, customer premise equipment, applications, and above all the economics
come together, then voice traffic will indeed be eclipsed. Based on Probe's ongoing analysis of
bandwidth to the home and small office, and all of the associated terminal equipment, standards,
and other issues surrounding the take off of residential multimedia, we remain convinced that
mass market adoption is at least five years off even in the best of circumstances (e.g., in the
heart of Silicon Valley). This is not to say that there are not already many households that
generate more data than voice traffic (e.g., in Silicon Valley), but rather that it will be some time
before early adopters give way to mass market consumers. For most communities, mass market
adoption is even further away (7 years or more).
Emergence of the Internet
The overwhelming driver of data traffic today is the Internet, as well as wide area private/virtual
private IP networks (e.g., Intranets and Extranets). While global voice traffic continues to grow
at less then 8% annually, data traffic is growing at roughly 35%. Although the Internet Protocol
has been defined for decades, its use became widespread only recently.
The market for Internet services was born in great part due to the privatization and
commercialization of NSFnet. Indeed several key players, including BBN Planet, MCI, Sprint,
etc., originally became involved in the Internet as government contractors. Others, including
UUNet and PSINet, launched commercial services in 1991, as soon as the NSF allowed it, and
were founded and staffed by NSFNet veterans. While proprietary online services and Bulletin
Board Systems (BBS') attracted aficionados and hobbyists for over a decade, the audience for
these services remained limited due to constraints on content, incompatibility between networks,
and exceedingly slow access, as well as the basic issue of limited PC deployment. In the early
1990s, however, several factors coincided to cause the Internet to become a viable commercial
platform:
1991 - NSF allowed commercial use of the Internet, and the first Commercial Internet
Exchange was launched;
1992 - The number of Internet hosts reached 1 million (largely academic and military);
1993 - WWW and the Mosaic browser were introduced; and
1994 - The penetration rate of PCs in US homes reached 33%; 28.8 Kbps modem standard
was approved.
By year end 1998, PC penetration amongst US households is approaching 45%. Importantly,
this figure takes into account households with multiple PCs, and excludes those with inactive
PCs.
FIGURE 2: INTERNET TRAFFIC: 1991-1997
Global
(Terabytes/Month) Internet
Users pass
3,000 50M
OSP
business
2,500 PC Penetration
model
in US Homes
fails; OSPs
reaches 33%; Internet
2,000 become
28.8Kbps Users ISPs
Modem Pass
1,500 Standard 10M
Approved
1,000 WWW and
NSF Allows
Mosaic
Commercial
500 Introduced
Internet Use
-
1991 1992 1993 1994 1995 1996 1997
Sourc e: NSFNet ac tuals to 12/94; Pr obe es timates after 12/94
By year end 1998, there will be an estimated 100 million Internet users worldwide (including
business, consumer, students, etc.). See Figure 3. For perspective, there are well over 600
million telephone users worldwide. Nonetheless, the Internet has proven to be the clear
technology winner as a user interface to a global public data network.
Figure 3: Worldwide Internet Users: 1997-2002
300,000,000
250,000,000
Internet Users
200,000,000
150,000,000
100,000,000
50,000,000
-
1997 1998 1999 2000 2001 2002
Europe/Middle East/Africa Asia/Pacific
North America Latin America/Caribbean
Source: Probe Research, Inc. estimates
Internet Telephony - Market Evolution and Drivers
In its brief life, IPT has already seen two distinct phases of market evolution, and is on the verge
of a third. In this section we describe the expected development of IPT as a market, and explore
the factors that will drive each phase of adoption of the technology. Figure 4 provides a general
overview of the evolution, including Probe's projections of global voice and fax over IP in
minutes of use.
Figure 4: GLOBAL VOICE/FAX OVER IP TRAFFIC
AND MARKET DRIVERS
90000.0000
80000.0000
Cost Parity
70000.0000
Millions of Minutes
60000.0000
New Applications
50000.0000
40000.0000
Feature
30000.0000 Parity
Tariff
20000.0000
Arbitrage
10000.0000 Hobbyist
0.0000
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Source: Probe Research, Inc.
Hobbyist Phase - IP Telephony as Ham Radio Substitute
In late 1995, a few technologists and hobbyists recognized the potential of the Internet to carry
voice traffic. Companies such as VocalTec, NetSpeak, Quarterdeck, and others developed a new
class of software that enabled any two Internet users to talk to each other from their PCs. This
approach, which relied entirely on the PCs at each end, required not only soundcards and high
speed modems at each end, but also that end users use the same software. Each solution was
proprietary – a VocalTec user could not talk to a NetSpeak user. Nonetheless, the draw of free
phone calls was compelling. By year end 1996, over a million copies of this class of software
had shipped, and players like Microsoft and Intel had jumped in to develop standards-based
client software as part of their larger vision of using the Internet as an alternative voice and
conferencing platform.
Not surprisingly, early adopters of this technology largely treated it as a chat room alternative.
VocalTec and other companies set up Internet relay chat rooms (IRCs), where users could go to
find other users with the same software, or where users might set up private meetings. Despite
all the awkwardness of this form of communication, literally millions of conversations were
completed between PC users over the Internet in 1996.
Tariff Arbitrage - Calling Card Replacement
The long distance communications market is extremely price elastic. This is important because
it is currently the underlying market driver for Internet telephony, and will remain so for the
next three to five years. To support this statement, consider the impact of competition on the US
long distance market in the 1980s. Figure 5 shows indices of price per minute versus minutes of
use of interstate long distance in the US during the period 1981 to 1993. As prices fell, traffic
grew considerably. Indeed the highest rates of traffic growth occurred during the period 1984
through 1988, when prices were falling their fastest (an average of 7.3% annually).
Figure 5: Price Sensitivity in Long Distance
18.0% 2.0%
Annual Percentage Change in Price per Minute
16.0%
0.0%
Annual Percetnage Increase in Minutes of
Traffic
14.0%
-2.0%
12.0%
Conversation
10.0% -4.0%
8.0% -6.0%
6.0% Price
-8.0%
4.0%
-10.0%
2.0%
0.0% -12.0%
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
Source: FCC and Probe Research, Inc. estimates
Early adopters of IP telephony quickly realized that the initial acceptance of the technology
hinged on its ability to bypass regulatory regimes and tariff structures. International call back
service providers, such as IDT, were among those most threatened. Their customers are
amongst the most price sensitive, and are willing to dial additional digits or suffer other
inconveniences to save money. Consequently, these types of service providers were the first to
develop IP telephony services. Other service providers, such as AT&T, Sprint, etc. also paid
early attention to IPT as a threat and therefore also an opportunity.
Simultaneously, vendors such as VocalTec, Vienna Systems, and Clarent Corp. recognized that
while the PC hobbyist market was a nice start, the installed base of multimedia computers was
rather limited compared to the installed base of traditional telephones. Moreover, the biggest
problem with IPT, the quality of the voice, was seriously affected by the end user’s own ability
to configure their system (the PC, the software settings, etc), and the reality is that non-hobbyist
users are more interested in communicating than in understanding the finer points of codecs.
The market potential of phone-to-phone and PC-to-phone IPT drove the development of first
generation of IPT gateways. And service provider fear of lost revenues provided the demand
that has fueled the next generation of IP telephony platform, the gateway.
IPT gateways, which enable end users to make IP calls from traditional wireline or wireless
telephones, can be deployed in an end user or service provider network. The function of these
gateways is to packetize the voice signal on the originating end of a call so that it can be
transported over an IP network. At the terminating end of the call, another gateway converts the
packets back, and re-assembles the voice signals for the traditional telephone network.
In the past year, voice over IP has eclipsed the Internet itself as a source of hype. There is little
question, however, that this market is no longer a vague fantasy shared by a few hobbyists and
technologists. Several service providers now report traffic in the millions of minutes per month,
and many more are ramping up trial networks as well as commercial networks globally.
There are a variety of factors driving growth in the internet telephony and fax market. Amongst
traditional service providers/PTTs, voice and fax over IP has caught on the fastest driven largely
by fear. Those with the most to lose are understandably the most aggressive in testing and
deploying new technology.
ISPs are grappling with growing competition and declining margins in the plain vanilla internet
access market. Consequently, most are exploring a variety of value-added services to increase
revenue per subscriber. Now that the technology is maturing, and large carriers have moved
beyond the announcement stage into the full-scale trial stage, ISPs are testing the water and
beginning to understand the true nature of the opportunity.
Tariff Arbitrage - Enterprise Trunk Replacement
Amongst enterprises, IPT networks have been slower to take off. This reflects several factors,
including lack of properly packaged solutions for the enterprise market, skepticism about the
quality of service afforded by IPT, and the immaturity of the equipment and software.
Enterprises have voice networks that work today, and need compelling business cases to replace
those. Moreover, the MIS and telecom purchasing decision makers still tend to be separate and
often at war. Nonetheless, enterprises have deployed thousands of IPT gateway systems, and are
also trialing IPT services in growing numbers. For many, IPT services are the point of entry --
rather than replace or overhaul the PBX, an enterprise might introduce cost efficiencies by using
IPT calling cards for traveling execs to call the home office for example.
All in all, however, the market is largely a service provider market today, and is expected to
continue to be for the next couple of years. The tariff arbitrage opportunity is, however,
expected to wane. As it does, we anticipate that feature parity and new applications will sustain
the next wave of market growth.
Feature Parity
The next wave of potential users of IPT that are attracted to lower prices will adopt the
technology only after it is equally convenient to traditional voice services. This phase requires
the next generation of IPT equipment. Although the gateway market has evolved at a
tremendous pace, the IPT gateways that are available as of late 1998 are still lacking in several
regards. Today's gateways are not interoperable with each other, they do not interoperate
effectively with the existing circuit switched network, they do not scale well, and they are
extremely expensive when compared to the traditional public switched network. However,
network equipment vendors are working feverishly on these problems. Probe estimates that the
number of engineers and developers devoted full-time to IPT has grown fivefold in the past 18
months. We expect to see substantial progress on all of these issues over the next 18 months.
Importantly, we are already seeing the beginning of feature parity, and this is what will drive the
next wave of IPT adoption by end users.
In early 1999, several service providers will introduce IPT and fax services that no longer
require the end user to dial any differently than they would for a regular phone call -- this is
referred to as single stage dialing. This and other key features that users expect of the traditional
voice network will be available in IPT networks as vendors introduce SS7 integration into their
IPT solutions. As was true in the US with the introduction of long distance services, it was not
until users had the option to do single stage dialing for their choice of carriers that competition
took off.
New Applications
While price is the key driver of the IPT market today, most observers agree that extending end
user feature control and functionality are the real benefit that will drive the IPT market in the
longer run. Indeed the very future of IPT depends in part on the development of compelling
applications. There are a variety of potential applications, many of which we have probably not
even conceived of, that leverage the integration of voice and data on a single network and user
interface. Internet telephony may well prove to be first killer application that finally drives
widespread adoption of computer based telephony.
Some of the applications that are already possible today include:
Multimedia Conferencing;
Multicast;
Collaborative Workgroup applications;
Call Center applications;
Message unification; and
IP Call Waiting.
Conferencing Applications
The first three are closely related. VocalTec, Nortel, and others sell software applications today,
generally H.323 compliant, that enable end users to simultaneously talk and share data over a
single IP connection, conduct videoconferences over IP, and share documents over that same
connection. Visual interaction and the ability to work with a document despite physical distance
is of tremendous value to certain applications. For example, a commercial building project in
Mexico City may require intermittent real-time interaction between an architectural team based
in New York and a builder team on location. Today, the two teams might pass faxes or
electronic copies of documents back and forth in non-real time between telephone and email
discussions of alterations that are required. With videoconferencing, both teams can make a
series of decisions in real-time as they simultaneously look at the plans and redraw them. This
might save literally days of elapsed time, and significant sums of money for a given project.
Other examples abound. Needless to say, there is also a significant residential/consumer demand
for videoconferencing as the cost becomes reasonable, the quality improves, and the
implementation becomes standardized.
Audio-conferencing, not to mention videoconferencing, remains one of the most expensive
carrier services. The highest cost of provisioning conference services is in the administration/
management and call set-up. Enabling users to do this work themselves via software has
allowed even traditional carriers to sharply lower their pricing. Conferencing is a prime target
for IPT, both due to dramatic pricing differences, and to the increased user control inherent to
IPT versus traditional teleconferencing.
One of the increasingly most common applications of audio conferencing is to make
announcements to various segments of the public -- quarterly reports to financial analysts, new
product announcements to user groups, etc. For this application, Internet solutions are especially
valuable as they enable users to visually follow along (the usual method now is to fax or email
ahead copies of presentation materials).
Call Center Applications
IPT offers compelling benefits in call center applications. The most common IPT call center
solutions enable a user to talk to an agent over the same line with which the user is browsing the
Web. This gives the call center agent the advantage of being able to view the users’ activities.
This is especially powerful in help desk and sales applications. Consider the numerous
occasions during which the help desk agent’s ability to observe a user typing commands would
save untold frustration and time, e.g., installing a new modem and struggling with archaic
communication set-up strings; trying to import and export documents from Windows to
Macintosh; installing any software; handling the inevitable bugs that result from marginal
compatibility, etc.
From an enterprise perspective, imagine a user browsing your web site. You sell insurance and
offer a comlpex assortment of options to be competitive. A variety of questions come to mind as
your prospect compares your prices to his current provider over his lunch break. The user clicks
on a button that dials through to an agent who is immediately armed with information about the
user, including the most recently browsed items, and perhaps additional information from a
personal profile form. The agent can make suggestions based on which items the user looks at,
and can guide the customer through decisions. Above all, you did not lose the prospect's interest
-- he did not have to go offline and place a call, nor was there any other delay.
Eventually, as Internet bandwidth and technology permit, videoconferencing with a customer
will be viable over the net. While this is of course possible, and practiced by a few bleeding
edge users today, it remains fairly impractical. Nonetheless, the time will come when video over
the Internet is feasible, and this will open a whole new set of possibilities.
IP Messaging
Voice messaging over the net offers several advantages, not just cost reduction. First, it is easy
to forward or broadcast a voice message over a network (whether its the Internet, or an Intranet,
or just a traditional private network). If a sales manager needs to get a quick message to the
sales force, and she doesn’t have her computer with her, a product called AMAIL enables her to
use a regular phone to send an audio message over the net via her corporate web page (be it inter
or Intranet) to the entire sales force with a single phone call.
Perhaps even more exciting is the hope of unified messaging. For the business traveler, this is
an especially important offering. Consider the activities of the average business traveler trying
to keep in touch from the road, repeating the following process 2 to 3 times each day. Check the
corporate e-mail, pager, voicemail, and a fax service or, more commonly, a helping hand who
will forward faxes to the user’s hotel room or other temporary space. A key application for
Intranets will be unified messaging, where a user can pull down all of these from the corporate
web page.
IP Call Waiting
Replication of traditional PSTN-based services in the IP network is a increasingly perceived as
key to the widespread adoption of IPT. Clearly, to achieve feature parity, a user must be able to
get all of the features they are accustomed to from the new network that were provided by the
old network. The benefit and promise of IPT is that new user interfaces will greatly enrich
many of the old standby services. IP Call Waiting is a fabulous example of an IP version of a
standard public network service that is richer and more powerful for the end user.
In the traditional PSTN environment, call waiting means that as an end user if I am on the phone
and another call comes in, I hear a beep, and I have the option to put my current call on hold and
answer the incoming call, or I can ignore the incoming call. Best case scenario, my service
provider may offer me the option of sending the call automatically to voicemail. With IPT Call
Waiting in a computer telephony world, I might be conversing with my manager on my web
phone (which has a browser up), and I can visually see that a call is coming in from an important
client -- their name appears on my screen due to the magic of caller ID. I put my manager on
hold -- he's calling from the airport -- and I answer my client's call to see what he needs. While I
am on that call, I can see my husband calling (because I can get more than two calls now), so I
send him to voicemail. In today's PSTN scenario, my husband would get a busy signal, and
wouldn't be able to leave me the message that he can’t pick up our daughter after school.
Note that this could all be occurring over a mobile web phone with a touch pad or voice input in
my car just as easily as over a PC at my desk.
Cost Parity
If IPT is ever really going to take off, however, there must be true cost parity between packet
voice and circuit-switched voice technologies. Today, the cheapest large scale IPT networks are
still roughly four times as expensive as their circuit-switched cousins. Network service
providers are pushing vendors extremely hard to bring the costs down. At the same time,
traditional equipment vendors continue to apply Moore's Law to circuit switching, and know
they must sustain their cost advantage as long as possible.
In theory operating an integrated voice/data network should be cheaper than managing two
separate overlay networks. In practice, however, this is highly debatable, and no one that has
really delved into the issue has been able to prove the business case, particularly for all of the
many service providers that have embedded networks. In the case of a new or greenfield
network, the business case has to assume either that the network deployment costs can reach
parity as networks scale over the next few years, or that the operational savings of IP are
significant enough to justify added network investment.
Probe expects that network equipment and operational cost parity will be achieved by the year
2002, and will be the primary driver of the IPT market over the long term when combined with
application, network management, and operational support flexibility. In the very long run, this
will drive replacement of circuit switching in the long distance network globally, as well as in
elements of the local network.
Market and Technology Barriers to Internet Telephony and the New
Public Network
As of late 1998, there remain several critical barriers to deployment of IP telephony, including:
• Lack of interoperability not only between IP telephony networks, but also between IP
telephony and the traditional circuit-switched network and other data networks;
• Quality of Service (QoS) and perceptions about QoS;
• Cost to deploy IPT;
• Lack of carrier grade network management, operational support, and billing/rating solutions.
Driven by significant pressure from service providers, a variety of vendors are accelerating
solutions and standards to address these barriers, and we expect to see most if not all of these
issues resolved within 18-24 months (i.e., by mid to late 2000).
Interoperability Challenges
Interoperability at a variety of levels presents the most serious set of challenges, and will require
a market environment that continues to push vendors extremely hard with the promise of real
rewards. There are several interoperability issues that need to be resolved. Most obviously,
network vendors must supply gateways and gatekeepers that are interoperable. No service
provider will be comfortable in the long run deploying a proprietary solution in large scale.
Gateway interoperability, which is not fully addressed by the H.323 protocol, is also expected to
be furthered by additional standards efforts such as IPDC. While H.323 specifies some of the
elements of interoperability, there remain several features of gateways that cause them to have
trouble interoperating.
New approaches, such as the proposed IPDC protocol, provide significantly increased gateway
interoperability as well as defining preliminary gatekeeper to public switched network
interworking. In addition, virtually every major gateway vendor is deeply engaged in
developing SS7 integration into their IP telephony solutions to ensure PSTN interworking, a
critical first step in the move to a public IP telephony network. Finally, vendors must address
the larger issue of seamless voice transport across IP, Frame Relay, and ATM networks.
Nonetheless, the standards process can be excruciatingly slow, dampened by in-fighting over
technology and intellectual property rights. The standards process in the IP space in particular
has become increasingly like that in the public telecom industry as each vendor’s commercial
interests and agenda take on growing prominence. Thus, it would be naïve to assume that
interoperability solutions are all on the verge of arriving. Rather, we expect to see gateway
interoperability almost immediately, followed by huge battles over gatekeeper interoperability,
since that is where the real network value and intelligence resides.
Quality of Service
In the early days of IPT, sound quality was a huge and overriding problem. In part this can be
attributed to the fact that most calls were PC-to-PC, and therefore required the end user on each
side to have proper equipment, and to optimize their software for each individual call. The
quality of a voice conversation over the Internet today is still highly variable, and often
unsatisfactory in comparison to a circuit-switched voice conversation. The quality of sound is
most impacted by latency (or delay) and lost/dropped packets.
The Internet is not a controllable environment. Consequently, as soon as an IPT call goes onto
the Internet, rather than a single controlled IP network, it is subject to unpredictable quality
issues. When an AOL end user sends a message to a PSINet customer, the message may travel
over only two networks (UUNET and PSINet), but it may also traverse others depending on the
routing tables, peering agreements, and other factors. Moreover, the first packet may travel very
efficiently while the next packet travels through ten times as many routers. The nature of the IP
protocol is that it offers connectionless networking. It was intentionally designed this way for
reliability and other reasons. The circuit switched network is connection oriented -- when User
A calls User B, the network sets up an end to end connection between those two points, which
ensures that all of the voice traffic is transported over the same route at the same rate.
The reality today is that IPT in a controlled network is toll quality. Indeed, calls over the
Internet using IP gateways are also often nearly toll quality. In the near term, most service
providers are really offering voice over an Intranet (meaning their own IP network) as a means
of controlling network quality. Enterprises are deploying IPT similarly over their own Intranets
initially for internal communications purposes rather than trying to interface with customers over
the Internet.
Over the next several years, service providers plan to deploy differentiated service qualities on
their IP networks for a variety of reasons, including to offer toll quality voice. Vendors and
service providers are developing a variety of standards and approaches that enable connection-
oriented services to retain higher priority over the connectionless IP network. RSVP
(ReSerVation bandwidth Protocol), Diff Serv (Differentiated Services), and approaches that
enable ATM/IP interoperability such as MPLS are all moving forward. The real story about IPT
in 1998 was the fact that quality is no longer the huge barrier that it was in 1997. By the year
2000, it will not be a major topic of discussion. However, the price of establishing quality of
service on IP networks that meets the requirements of toll quality voice may well be that the
economics of IP lose to the economics of circuit switching for years to come. If deploying IPT
means establishing permanent connections over an IP network, the economic advantages of
connectionless transport are lost, and a host of network management and bandwidth issues raise
their heads. Fundamentally, the economic case for IP networking becomes less clear.
Cost to Deploy IPT
From a service provider perspective, deploying IPT gateways is much more expensive today
than deploying comparable voice switching equipment on a per port basis (where a port is a
measurement of potential simultaneous calls). IPT gateways sold for an average price of over
$1,800 per port in 1997 and will be down to about $600/port in 1999 for a networked, managed
port. The same port in a traditional carrier network would cost less than $100. Thus traditional
carriers continue to have objections to large scale deployments of IPT in lieu of circuit
switching.
The cost of IPT gateways is declining very rapidly, and no one has yet developed a meaningful
analysis of the loaded cost of operating and maintaining a circuit switched voice network versus
an IPT network. As we discussed in the previous section, the true cost of an IPT network with
quality of service, billing mechanisms, and all the other paraphernalia of the traditional PSTN
added in may well be greater than the cost of today's PSTN for the same capacity. Unless the IP
model is proven to be lower cost in the long run, we will not see the fourth and most promising
stage of IPT -- cost parity -- take off.
Operational Support Systems
Finally, today's IPT solutions do not have the associated network management and operational
support systems developed to the same sophistication as the PSTN. We expect this market
barrier to be alleviated in 1999 and 2000. A variety of software vendors, from Bellcore to Portal
and beyond, are applying tremendous energy and resources to the development of IP-oriented
network management and OSS as well as billing systems. However, these are clearly a key
requirements of any service provider that hopes to scale up IPT as a major offer.
Regulation - The Very Real Threat
There is a critical race in progress today that has serious implications for the entire IPT industry.
This is the race between a shifting regulatory environment and the elimination of technology
barriers to IPT. Today, the entire IP telephony industry is being funded and driven by a tariff
arbitrage business case as we described above. Either of two opposing forces could throw a
wrench in the development of the IPT market. If either scenario plays out, then feature parity
and compelling enhanced applications become much more urgent.
In the first instance, if pricing of the PSTN reflected true costs rather than government subsidies,
the tariff arbitrage demand would disappear, and IPT would have to offer lower true cost or
sufficiently compelling enhanced applications to drive demand. For the past decade, much of
the world has been moving toward deregulation of the telecom sector. Many government owned
PTTs have been privatized. The World Trade Organization (WTO) agreed on a plan to reduce
international settlement rates globally, and the cost of telecom has been declining for much of
the world for several years now. The reduction of international telecom costs in particular eats
into the demand for IPT as long as IPT is driven by the argument that it is cheaper. The general
consensus has been that this movement of tariffs and settlement rates downward will continue to
occur at a reasonably slow pace, leaving sufficient room for IPT providers and vendors to fund
the fledgling market in the near term.
However, one should never forget that most IPT service provider business plans today rely on
their ability to offer IPT at essentially a subsidized rate, since IP is not subject to all of the same
tariffs as PSTN. Even in the US, as soon as companies like Qwest and Netcom lose their
enhanced services provider exemption, which currently enables them to avoid paying access
charges that AT&T and other long distance carriers pay to local carriers, their ability to offer
IPT any more cheaply than AT&T offers PSTN will disappear.
There looms a different scenario, however, that has not garnered the notice it probably should,
and that is the specter of a reversal in global trends towards deregulation. As with all
progressive movements, telecom deregulation is not universally popular. Indeed, deregulation
itself is a highly politicized process. There are in every country many interests, from labor
unions to local copper cable suppliers, that would prefer a status quo or more regulated telecom
industry structure. Understandably, shareholders in British Telecom would prefer to have
maintained monopoly status rather than face competition. Many governments benefit from
international settlement rates, and are strongly opposed to the FCC's proposal for globalization
of international carrier charges. If these forces were to gain strength for any myriad of reasons
over the next few years, a serious disruption of the IPT market is possible.
Indeed, IPT is outlawed in many countries, and is illegal but currently tolerated in others. One
can certainly construct any number of political scenarios that might turn the tide towards more
closed rather than open telecom sectors around the world. In particular, there is the very real
threat of a global recession as the economies of Japan, Russia, and most of Asia suffer from their
respective financial crises. In times of economic uncertainty, we often see a shift in the political
environment, and we almost always see a return to nationalism. Consider the recent election in
Germany, which replaced an international leader with a leader much more focused on the
domestic agenda, one who is moreover closely backed by labor unions. In reality, telecom
policy is continually subject to shifting governments and business interests. We should not
assume that the environment of the very recent past is a foregone conclusion for the future.
Service Market Opportunities in Internet Telephony
There are a variety of market opportunities generated by IPT. In general, we segment these into
three major service markets: Wholesale (or carrier's carrier); consumer/small business; and
enterprise/institution. Figure 6 below provides an overview of Probe's projections of global fax
and voice over IP service revenues in each segment over the next six years. These figures do not
take into account PC-to-PC IPT, as that activity generates no direct service revenues.
As with all carrier service markets, any element that is sold as a retail offer may also be provided
by a wholesale carrier to service providers with limited or even no facilities. Today's IPT
market is characterized by a great deal of calling card activity, which reflects the fact both that
multi-stage dialing is required, and the target customers are very price focused. The way many
calling card sales work is that a retailer buys minutes from a carrier that sells them those minutes
in packs of cards. An increasingly common approach is to sell pre-paid calling cards at retail
outlets such as office supply stores, convenience store, etc. The wholesale market shown below
is therefore money that comes from the retail sector -- thus there is an inherent double count of
revenues in the figure below that one can adjust simply by pulling out the wholesale. Note that
wholesale services is only counted as the revenues that a carrier receives for wholesaling a voice
or fax over IP solution -- we are not including for example, the revenue that a backbone ISP
generates by providing pure backbone services to another carrier that then sells a total IPT
solution using that backbone. The wholesale market is a significant component of the total IPT
market, and will remain important as service providers seek to provide global solutions.
The residential and small business market currently dominates the retail sector for IPT, again
because the calling card model has taken off so successfully for the moment. Over the next five
years, this segment is expected to remain dominant, with an increasingly strong business
element. Large enterprises and institutions typically use private leased line networks or virtual
private networks for much of their voice communication. As IP-based VPN's become more
widely deployed, we expect to see increasing us of IPT over VPNs by larger organizations.
However, migration will be very slow, due to the resistance that always exists to replacing
working systems, and we expect most of the large enterprise and institutional market to be
driven by expansion, new branch offices, etc.
The retail markets will also be driven more over the long run by enhanced applications. Again,
we expect to see these take off amongst consumer, small office/home office (SOHO) and small
business first. Some service providers, notably AT&T, are developing Internet-centric
applications that integrate traditional internet communications (such as chat and email) with
voice targeted at consumer applications. Naturally the hope amongst service providers is to
create value added services that raise IPT from a cheap PSTN substitute to a premium service
offering feature rich communication.
Figure 6: IP Voice/Fax Service Markets by
Segment
14,000
12,000
10,000
$US Millions
8,000
6,000
4,000
2,000
0
1997 1998 1999 2000 2001 2002 2003 2004 2005
Consumer & Small Business Enterprise & Institutions Wholesale
The Value of Early Market Leadership/ Today’s Internet Telephony
Service Providers
In 1998, the IP telephony services market is dominated by a small handful of pioneers, including
Delta Three, IDT, AT&T, ITXC, and VIP Calling. Others, such as Qwest, PSINet, KDD,
Deutsche Telekom, and ICG/Netcom are only beginning to sell services in any meaningful way
in the 4th quarter of the year. Interestingly, large ISPs such as UUNET and PSINet have
generally shied away from full IPT, in part we believe due to fear of bringing regulatory
attention to the Enhanced Service Provider exemption in the US that protects them from paying
access charges. UUNET has focused on IP-based fax services, while PSINet is rolling out VPN
based services that are much less vulnerable to the argument that they are just selling voice
minutes.
The three current market leaders globally are IDT, AT&T, and Delta Three, with VIP Calling a
not too distant fourth. Market leadership is difficult to establish -- in such a fast growing market
we often find that it changes month by month. All four of these are, however, carrying millions
of minutes each of IPT on their networks. Despite the entry of other players, Delta Three and
IDT have managed to retain leadership by leveraging their early market experience. Perhaps