"We started out with the Net running on top of the phone system, and we'll end up with telephony running on top of the Net. A completely unregulated network is toppling its highly regulated predecessor. That's extraordinary." - Eric Schmidt, CEO of Novell
The Internet versus the Public Switched Telephone Network
A central office complex of a large telephone company is a humbling place. A central office is one of the visible spots at which the shared telecommunications infrastructure is managed by a controlling entity. The amount of information, technology and engineering data that must coalesce into a seamless transaction is staggering. Just keeping track of what wires go where and what telephony services they carry is a large-scale data engineering problem involving tens of millions of records. Those wires, of course, go from customers to switches and computers, and ultimately to other telephone customers. The wires are capable of comprising circuits that allow more or less completely reliable, universal and commercially sustainable voice and data communications. The confluence of communication, data processing and business processes that must be established each time a phone call is made is, in fact, so complex that to the outsider the mere feasibility of phone calls seems improbable. A central office complex epitomizes the Public Switched Telephone Network (PSTN).
On the other hand, there is the Internet. It is not universal, not very reliable and is not controlled by any particular entity. The Internet has nothing comparable to a central office. The only things that are needed to establish service are access to the network and a computer that is aware of the Internet Protocol (IP) used to break messages into packets, route them to destinations and reassemble the packets into something like the original message.
However, it has been possible for several years to bypass the PSTN using the Internet. It started with freeware called CU-SeeMe developed at Cornell University. Two years ago "Net phones" began appearing in greater numbers. Net phones are PC-based applications that handle the digitization and compression of real-time, two-way voice data of the kind that would be needed to carry on a conversation. Virtually everyone who tries Voice-over-the-Internet applications complains that the technology is really only suitable for hobbyists. Voice quality is usually a tinny, walkie-talkie parody of telephone-grade communications. Service is not universal, of course, since not everyone that you want to call has a PC. It is also not reliable and it is not uncommon that Internet traffic congestion or other problems isolate groups of potential telephone users in islands that are unreachable and without service. Many essential services are not available to users of Net phones. And, most significantly, most Net phones cannot receive "real" telephone calls made by users of the PSTN.
There are more than 650 million phones connected to the PSTN. The best estimates are that within three years there will be 100 million host computers connected to the Internet. The entire worldwide capacity of the Internet is smaller than the PSTN by several orders of magnitude. It therefore seems unlikely that Net Phones will somehow morph into "Internet telephony" in a way that competes with or threatens the PSTN. Indeed, even the most optimistic estimates predict that worldwide Internet telephony revenues may reach $2 billion by 2002. Compare that with the $475 billion spent on basic telephony in 1996 - clearly Internet telephony will within five years be only a small bit player in the global telephony marketplace. Why, then, is it one of the hottest topics among Internet technologists and why would people predict that Internet telephony will "topple" the PSTN and become the next telephone network? The answer that is emerging involves the evolution and interaction of three forces.
PSTN Economics
The PSTN is an end-to-end, circuit switched network. When there is a mixture of data and voice traffic, network operators can define and price both access and services to provide acceptable tradeoffs. However, data traffic now dominates voice traffic on the PSTN and in order to handle all of the bits that must be transported, much of the data traffic will be transferred to the Internet, which is a general-purpose data network. How that will be done is a subject of intense debate but the outcome is not in doubt. In the end, the Internet will carry a great deal of data traffic. It will also carry some voice traffic (as data). But as the imbalance becomes more pronounced, the voice traffic and, most importantly, the services enabled by that traffic, will become a commodity and will be priced accordingly.
Internet Growth Curves
Everything about the Internet is on an exponential growth curve. The amount of traffic, revenues from Web advertising and the number of connected hosts all double with certain regularity. That means that the greatest opportunities for commercial growth lie on the Internet. Commodities, on the other hand, are more likely to be subjected to the damping supply-demand effects of a stable market place.
The Telecommunications Reform Act of 1996
Telecom reform was intended to open telecommunication in the U.S. to increased competition. The extent to which it will succeed is not clear at this point. One thing is clear, however. It has created an entire underclass of competing "carriers" - companies that are not at present carrying substantial amounts of voice traffic but will develop the capabilities to do so in the future, provided only that adequate transport facilities are available.
In short, since the Internet is not an end-to-end network, many new businesses will spring up to exploit exponential growth. Many of these businesses will look very much like the application software business because the Internet makes this the preferred way to provide services. If new carriers provide application software that provides all of the services of the PSTN at reduced or competitive prices they will draw those services into the Internet at the expense of the PSTN, essentially relegating the existing telephone network to carrying bits only. Since telecom reform requires the current (or "incumbent") carriers to open their networks to a broad spectrum of the new competing carriers, it will be increasingly difficult for the incumbents to jump on the Internet growth curve.
How fragile is the economic model underlying the PSTN? The answer lies in the dynamics of the long distance market. Peter Sommerer, president of Newbridge Communications, did a back-of-the-envelope estimate of the likely impact of Internet telephony on long distance carriers: "Large U.S. long distance carriers make around 6% net profit each year. To remove this profit, 1 in every 17 phone calls needs to be made over VOI [voice-over-the-Internet], or 5.7 million households in the U.S. need to switch . . . " Is this achievable? Prior to 1996, the market was subject to considerable churn. Approximately 25 million customers or 25% of the market in the U.S. would switch carriers each year. Deregulation and increased competition in the long distance market push this number even higher. So, if Internet telephony captures just 25% of the customers who are choosing new carriers, all profit will have been driven out of U.S. carriers' business.
What is Internet Telephony?
The raw capability for making a voice connection over the Internet is not the same thing as Internet telephony. In order to approach the reliable, highly available, service-rich capabilities of the PSTN, developers of Internet telephony technology must either duplicate or reinvent the functions of the voice network.
The PSTN consists of an analog voice network and a digital Common Channel Signaling Network that is used to route and set up calls and to make the database queries needed to establish intelligent network services like 800/888 calling, credit validation and local number portability. Customer-provisioned switching equipment initiates transactions with service control points. A service control point in turn handles routing, telephony and business functions. Eventually, a circuit with the right service characteristics is established between the caller's customer-provisioned switching equipment and the customer-provisioned switching equipment that is being called. One great advantage of the intelligent network services architecture is the management of network services. Without intelligent network services, adding an advanced service like call forwarding would require reprogramming the software in the many switches to which the service subscriber can be routed. With intelligent network services, the service provider can update a couple of centralized databases.
In an Internet telephony network, the Internet and intranet are indistinguishable from a technology standpoint since they are both IP computer networks. At a more fundamental level, however, there are vast differences. Intranets are closed networks. They are typically operated by a company or other enterprise as part of a centrally managed corporate infrastructure. The Internet is not. One way this difference becomes apparent is the relative abundance of "bandwidth" on corporate intranets. Another is the relative abundance of computers in the enterprise. In an intranet, issues like voice quality and universality become less important than for the Internet as a whole.
Gateway call servers play a pivotal role in Internet telephony networks. They provide, for example, all of the signal processing functions that convert between the digital packets of an IP network and the analog signals of the PSTN voice network. A call server maintains an IP connection on one side and a voice network connection on the other. Because the voice network is circuit switched, the size of the server is measured by the number of "lines" that it interfaces. A 120-line call server, for example, can maintain 120 simultaneous circuit connections. In a major carrier, thousand of lines are required for commercial viability so the architecture of call servers must be scaleable in the sense that the cost and performance of operation of a network with multiple servers must not increase in unpredictable ways. In particular, it should be possible to achieve economies of scale in a large-scale deployment of gateway call servers.
Call servers also provide many of the signaling and intelligent network services functions of the PSTN. Call routing and translations between phone numbers and Internet addresses are handled by the servers, as is registration of a PC that is online and ready to receive calls. Assuming that many services (perhaps even access itself) will be paid for, call servers are the ideal places in the Internet telephony network for logging, usage accounting and billing. For enterprise networks, call servers will be capable of firewall and security management. They will also be the platform for unified messaging, centralized database lookups, integrated call processing and advanced directory services like single number addressing.
Just as in the PSTN, intelligent network services for an Internet telephony network will help manage the complexity of peer-to-peer communication. To communicate, both the calling and the called parties must have compatible equipment. For PCs this means that they must both use the same voice encoding and compression programs. Because there is no real "dialtone" for the Internet, the called party must be "registered" to announce his existence to anyone who wants to call. This is sometimes bundled with other useful services. For example, Netscape Communicator's ConferenceTM program allows callers to compose and send voice mail messages to users who are not registered. This voice mail is packaged as a multimedia Internet message enclosure (MIME) and delivered as e-mail.
Telephone Calls on the Internet
A telephone call involves parties who need to establish communications. One of these is the originator or the calling party. The calling party initiates a sequence of events that result in the setup of the call, transmission of information and finally a process of freeing whatever network resources were committed to the call (in the PSTN this phase is referred to as tearing down the call). An Internet telephone call requires a substantial involvement of an IP network in transmission and possibly other phases of the call.
Non-Real-Time Communications
Non-real-time services are already common on the Internet. These services include fax, voice mail, paging and Personal Digital Assistants (or PDAs). The calling party uses either a PC or conventional equipment to initiate the call. Transmissions are stored on a server until the called party retrieves them. Such communications are not real-time because there is no real interaction between calling and called parties during transmission.
Real-Time Communications
PC to PC. Call originators and receivers announce their availability to receive voice calls by starting PC clients that communicate with servers to register the PCs on their respective networks. Part of this communication associates names with Internet addresses of the computers. Thereafter, the originator uses the registered name to signal the receiving computer user. Once the called party accepts the call, a series of communication protocols establishes the point-to-point connection that simulates a PSTN circuit.
PC to Phone. A calling party uses the local client application as an analog-to-digital converter for communication with a call server. The server handles all details of signaling (dialing) with the PSTN. The PSTN in turn manages call set-up, transmission and tear-down just as if the call had originated in the PSTN.
Phone to PC. A caller uses PSTN dialing to access an Internet Call Server on which the called party is registered and available to receive calls. The server uses the Internet telephony protocol stack to establish server-to-computer connections and manage the PSTN connection to the caller's phone.
Phone to Phone. Both originators and receivers use familiar PSTN telephones. To complete a call, the caller accesses an Internet Call Server which routes the connection over the Internet to a server that is "local" to the call party, at which point the call hops off the Internet to a local PSTN connection. In phone-to-phone calls, the Internet is used strictly as a bypass network, avoiding PSTN tariffs and tolls.
Internet Telephony Solutions
For some developers, Internet telephony offers a cheap alternative to conventional long-distance service by transporting phone calls over the Internet instead of using the facilities of the PSTN. Bypassing toll charges by using "free" Internet bandwidth is undoubtedly an attraction. However, most analysts believe that simple cost-avoidance will not sustain Internet telephony in the long-term. Rather, the main attraction will be the vastly greater potential of seamless integration of wireline, wireless and Internet network services in a universal, dependable communications network.
Internet telephony users will interact with services using familiar telephone handsets, multimedia-capable PCs, cellular or personal communications services phones, mobile or wearable devices, broadcast and cable-TV receivers, and other equipment that have not yet been invented. Moreover, these devices will be able to communicate with each other, a feat that is difficult to achieve if computer, voice, data, wireless and broadcast networks evolve independently.
Internet telephony users will have access to Internet versions of common and popular PSTN services:
Directory services
Call waiting and similar services
Calling card validation
Caller ID
Because computers offer processing power, and data networks are capable of high data rates for information exchange, Internet telephony also promises new services that are difficult or impossible to achieve otherwise:
Multimedia conferencing
Collaboration by geographically distributed work groups
Agent-enabled electronic secretaries to manage messages and schedule meetings
Intelligent software for speech recognition and performing routine tasks.
To benefit from a true end-to-end Internet telephony network, customers also need messaging and service capabilities. Call servers, messaging servers and service nodes will access and eventually interface with other network technologies such as the Intelligent Networking systems presently deployed in the PSTN.
For corporate customers, these call servers will also provide features to interface with (or maybe replace) Private Branch Exchange (PBX) systems, allowing enterprises to customize, control and manage calls.
For corporations, the initial attractiveness of Internet telephony is cost avoidance. However, in the long run the true value of Internet telephony lies in the potential for using value-added services to enhance employee productivity and effectiveness. Business users can control the quality of service by designing specialized routing strategies. For example, corporate or user-defined rules can be used to handle non-real-time voice mail in one manner and critical multimedia sales or training calls in another. Profiles can be defined for customization by the user to direct call to wireline phones, cellular handsets, secretarial pools, PCs, e-mail and FAX servers, pagers and call centers. A single number will suffice to identify uniquely an individual subscriber to all of these services.
Telephony requires access to intelligent network services. Increased reliance on 800/888 databases, line information databases for calling card transactions and other intelligent network technologies will place an increased demand on Internet telephony to provide the same services in a seamless way. Moreover, call servers will allow rapid access to customer services. Today, business services such as Switch Direct are used to redirect calls "en masse" to alternative locations. This is especially valuable during emergencies. Tomorrow, the same capabilities will be used to detect busy signals during call set-up and to redirect those calls to wherever the user happens to be, a sort of Reach-Me-Anywhere service.
Commercial Promise
As of today, there are two dozen independent Internet telephony networks in operation. Some of these networks, such as the several successful fax-over-the-Internet services, have global deployments, customer service operations, acquisition and billing services, and can be largely relied upon for replacing PSTN services. Most of the others, however, are not suitable for commercial use. The networks cannot communicate with each other, for example, and directories do not exist for establishing internetwork connectivity. Most importantly, few commercially viable services have been developed. They are at present toys for computer hobbyists.
Like the Internet growth curve itself, growth of Internet telephony will be dramatic once the initial networks establish customer bases and large networks operators (like corporate intranets, Internet service providers and telephone companies) begin to rely on the services. Whether or not Eric Schmidt's direst predictions materialize, the Internet will become a major voice network over the next couple of years. It may even be the next telephone network.
Richard A. DeMillo is vice president, Information Technology and Internet Applications, at Bellcore. [email protected]