Converged Networks and Services: Internetworking IP and the PSTNby Lawrence Gabuzda, Hui-Lan Lu, Lawrence R. Gabuzda
A much-needed guide through the thicket of conflicting IP telephony technologies.-Scott Bradner Convergence of the Internet and the Public Switched Telephone Network (PSTN) makes possible a wide range of new telecommunications services. Written by the industry leaders at Bell Laboratories, this book puts the future of convergence in perspective and helps you make
A much-needed guide through the thicket of conflicting IP telephony technologies.-Scott Bradner Convergence of the Internet and the Public Switched Telephone Network (PSTN) makes possible a wide range of new telecommunications services. Written by the industry leaders at Bell Laboratories, this book puts the future of convergence in perspective and helps you make decisions that leverage its power. The authors describe the network infrastructure, then explain how the PSTN and
Internet are being modified to work with each other. They also define and explore three aspects of the convergence: 1) interworking with VoIP, 2) IP-based service control,
and 3) remote access to IP networks. With this book, you'll get an insider's
perspective of key technologies, an in-depth review of the relevant standards, and a detailed survey of emerging applications and products.You'll also learn how to determine which products and services are right for your organization. Wiley Networking Council Books put technology into perspective for decision-makers who need an implementation strategy, a vendor and outsourcing strategy, and a product and design strategy. The Council of series advisors is composed of three of the the most influential leaders of the networking community:Visit our Web site at www.wiley.com/compbooks/
Visit the Networking Council Web site at www.wiley.com/networking council
Read an Excerpt
Note: The Figures and/or Tables mentioned in this sample chapter do not appear on the Web.
The Technology Is Relevant
So far, we have surveyed standards and technology, discussed customer environments and applications, and talked about how to match these to the range of product technologies currently available in the realm of Internet and telecommunications integration. If you have made it to this point and have discovered some product categories that address one or more of your pressing needs, you may already be a believer and need no further convincing that the technology is relevant. There is, however, quite an active debate in the industry about the degree to which all of the telecommunications applications that are currently satisfied by circuit-switching technology will migrate to the Internet as some form of voice over IP (VoIP).
Some in the industry take the visionary approach and cite a few stark facts that seem to get to a conclusion in shortcut fashion. The simplest analysis relies on comparing the amount of network traffic (by some consistent measure) that represents conventional voice with that which represents IP data. IP data traffic is growing very quickly, and conventional voice traffic is growing relatively slowly, implying an eventual crossover in traffic volumes. Some large carriers are able to estimate, based on their own measurements and customer information, when the crossover will occur for them. WorldCom, for example, has said it should occur some time during 2001 in the long-distance network it operates.
The reasoning proceeds as follows: Since IP data will eventually represent the larger traffic volume, it will be more economical to convert voice to IP and carry it that way, just as it made sense to carry data over the telephone network when the volume of data traffic was small compared with the volume of voice traffic (and, going back to our history lessons in Chapter 1, just as it made sense in the mid-twentieth century to multiplex the diminishing quantity of telegraph traffic onto telephone carrier circuits).
This reasoning appears fundamentally sound, and, if you are the CEO of a large traditional telephone carrier, it certainly provides enough motivation to get your research and long-range planning departments to take a hard look at migration to an IP-based infrastructure.
However, in reality, nothing is quite this simple--certainly not the economics of the carrier industry. Also, the very idea that all forms of communication will converge on one infrastructure that is based on one protocol, whether IP or anything else, sounds a bit naive when said aloud without the aid of a full-color animated PowerPoint presentation. After all, it's never happened before. Even the telephone network, at the height of its monopolistic, government-owned-and- regulated powers, did not succeed in eliminating all other forms of communications technology. How likely is it then to happen in a market-oriented environment driven by customer choice, where there will be ample rewards for a technology that provides even a small but customer-perceptible advantage over IP for a given application set?
It appears, therefore, that there may be a few more things to be said about whether the technology of the integration of the Internet and telecommunications is relevant, or to whom or for what or for how long. These considerations will be the subject of our concluding chapter. Among the topics to be discussed are the following:
- The cost structure of the carrier business--which elements are sensitive to protocol differences at the IP voice layer, and which are perhaps not sensitive to this at all?
- Life cycle costs from the end-users' viewpoint.
- Regulation--local, national, and international.
- The impact of multimedia.
- The requirements of specialized applications.
Economics of the Integration of the
Internet and Telecommunications
Carriers and enterprise networks differ to some degree in the way in which they are affected by network integration economics. Accordingly, in the following sections, we will begin by treating carriers and enterprise networks separately, and then follow up by looking at some joint effects.
To understand the economics of the integration of the Internet and telecommunications from the carrier point of view, it is necessary to delve into the cost structure of the carrier business and then to attempt to determine how it is impacted by the opportunities for integration at either the service or the infrastructure level. It should be said at the outset that although today one can find any number of articles in both trade journals and scholarly publications that discuss the economics of, for instance, Internet telephony, any such analysis is actually fraught with various fundamental difficulties. To begin with, the sources of information are necessarily somewhat indirect. Essentially, any carrier that has a very good understanding of its own cost structure will usually treat the details of this information as a closely guarded competitive secret; any carrier that does not understand its own cost structure is not going to be of much help. Public record information, such as that contained in annual reports and reports to the FCC, includes only gross-level cost breakdowns. Also, there are inherent uncertainties in estimating the costs of new types of networks that have not yet been built. These include not only uncertainties in the costs of new hardware and software, but also, perhaps more important, an unknown impact on operational costs.
Perhaps we can truncate our list of caveats and warnings here. You get the idea; the answers are not going to be obvious, and, indeed, this is part of what makes the industry so interesting and dynamic at the present time.
First, let us clarify that we are talking about costs to the carrier. There is, of course, an economic theory that says that prices are equal to marginal costs, but that theory assumes an ideal free-market environment rather different from the collection of regulated, unregulated, deregulated, competitive, monopolistic, government-owned, and private ISPs, CLECs, ILECs, IXCs, MSOs, Post, Telegraph, and Telephone (PTTs), former PTTs, and second, third, and fourth operators that constitute the carrier industry today. Differences in pricing, which are solely or primarily driven by regulation, will be the subject of a later section. Here we are trying to get at the impact (if any) of Internet-telecommunications technology itself on underlying costs as seen by the carriers.
A first step (and possibly the most helpful one of all) is to understand the categories of costs seen by a carrier providing voice, data, or multimedia services. Table 12.1, for example, shows the AT& T Annual Report for 1997 (a year that happened to be relatively free of extraordinary charges for restructuring or whatnot), which reveals the following breakdown of that U. S. interexchange carrier's operating expenses.
What is bundled into these categories? What effect could the integration of the Internet and telecommunications possibly have upon them? Let's take these four components in descending order of size.
Table 12.1 Components of Operating Expense
Source: Derived from AT& T Annual Report, 1997.
1. Access and other interconnection is mainly what AT& T pays to U. S. local exchange carriers (LECs) for the services of originating and completing the calls of end-user customers and for similar services such as providing the local ends of long-distance private-line circuits. Recall that AT& T's main business (and even more purely so in 1997, before its recent forays into the cable television market) has been providing the long-distance portion of phone calls and private line circuits. Listing this very large expense as a separate line item is, in part, AT& T's way of highlighting what a large fraction of its costs are actually determined by the LECs and the federal regulators who approve the level of carrier access charges. This expense category is unique in that it is not directly affected by technical changes to the network, such as the integration of Internet and telecommunications from a network or equipment perspective. Rather, it is determined by regulatory treatment. At the moment, if AT& T acts as an ISP in providing a service, access charges do not apply. Avery interesting question is whether this regulatory treatment would change if such an ISP began handling large volumes of telephone traffic as voice over IP.
2. Selling, general, and administrative (SG& A) expense bundles together most of the corporation's activities, other than running the network itself. Much of this has to do with the very high expenses associated with sales and marketing in the highly competitive U. S. long-distance telecommunications market. This includes paying for sales staffs to maintain relationships with large business customers and for advertising campaigns to woo smaller business customers and consumers. It is hard to foresee a direct, near-term impact on these costs from integrating telecommunications and the Internet. To the extent that the technology is used to provide the same services by another means, SG& A costs should remain about the same. To the extent that the integration of the Internet and telecommunications creates new and more complex service options, sales costs could increase somewhat in the near term.
3. Network and other communications services expense finally begins to get close to the network itself. Mainly, services are the cost of paying people to run the network. These people do things like testing and repairing switching and transmission equipment and performing the administrative tasks that are necessary when new equipment is added to the network, when new routes are established, or when new customer services are turned up.
What additional impact the integration of the Internet and telecommunications will have on these network operating costs is a very interesting and important question, but a tough one to answer. The mainstream thinking seems to be that network operational costs for an embedded carrier with an existing circuit-switched network will go up initially as its Internet business grows, since the carrier will have to run an extra network in addition to the ones it already owns, but that it will then go down (possibly significantly) as it becomes possible to carry all services over a common IP network. However, there are a number of uncertainties associated with this scenario.
To begin with, the operational cost differences between circuit-switched voice networks and IP networks applied to multiple services, including telephony, are not well understood. Telephone companies today operate their circuit-switched networks with the aid of operations support systems (OSSs), which are either supplied by the telecom equipment manufacturers, developed in house, or developed by third parties such as Telcordia. Some of these systems support administration and maintenance of network elements, such as switches, transmission systems (terminals, multiplexers, line systems), and the signal transfer points of the SS7 networks. Other systems support service-level functions, such as loading data into switches and service control points as customers subscribe to services as simple as call-forwarding or as complex as a nationwide or international toll-free service or virtual private networks.
At the network element level, a voice-over-IP service would involve operations support for new network elements. In addition to IP routers, these would include the packet voice gateway systems, which convert voice from circuit to packet form. Whether configuring such systems is simpler or less costly overall than performing the equivalent functions for circuit switches is, therefore, a key issue.
Operations experts who have studied proposals for large voice-over-IP networks have concluded that there are substantial opportunities for savings at the network management level due to easier bandwidth management (use of virtual trunk groups in place of physical trunk groups) and a probable reduction in the total number of network nodes. These effects are illustrated in Figures 12.1 and 12.2, which show trunking patterns in an interexchange carrier before and after the application of packet gateways. [The figures are highly simplified to make the point; an actual U. S. interexchange carrier, depending on its size, may have many more circuit switches (at least one per metropolitan area), and may or may not have full trunk group interconnection, depending on what circuit-switched routing algorithms are employed by the switches.]
At the service management level, differences may be fewer. Indeed, there may be incentives to utilize the same Intelligent Network infrastructure to provide, for example, toll-free services over an IP-based infrastructure, in which case the service management costs are probably identical in the two cases.
The real opportunity for dramatic operations costs savings with an IP-based infrastructure is, therefore, linked to the possibility of network consolidation. Telecommunications companies today typically operate several networks that are not fully integrated. There is the circuit-switched voice network. There is a private-line network, which shares the same transmission infrastructure but not the circuit switches, instead relying on digital cross-connect systems to manage connectivity. There may be multiple data networks providing ATM, frame relay, and IP services. Each of these networks, in turn, has its own unique set of OSSs and operations workforces.
Because circuit switching does not seem to be up to the job of handling the data side, it has seemed clear for some time that some form of packet technology would be needed as the basis if consolidation of these networks were undertaken. From the late 1980s through the mid-1990s, there was a strong consensus in the telecommunications industry that ATM would eventually enable consolidation. Even before this, in the early 1980s, AT& T devoted a considerable amount of study to an internally developed technology called wideband packet (a form of high-speed frame relay), in which potentially large operational savings from network consolidation were identified. The reason these earlier visions did not materialize is that faith in the operations savings studies was not strong enough to support the massive investment costs of an infrastructure change-out. What may be different about IP is that the very high growth rate of IP-based services may reduce the need for leaps of faith--a large IP-based network will come into being in any case, and it can then be used as the basis for integration.
4. Depreciation and amortization at last gets down to the tangible things that make up the network--the costs of hardware for switching and transmission. At 8.9 percent, it is the smallest of the operating expense categories, but 8.9 percent is certainly not a negligible portion of total expense. There are several ways in which the integration of the Internet and telecommunications may save on hardware costs:
- Savings from overall network consolidation
- Voice compression
- More favorable switch port costs
Network consolidation savings are the hardware equivalent of the potential operations savings for the converged networks described previously. To a great extent, these derive from the principle that larger networks are more efficient. When all traffic types are put on one switched network, instead of being split among multiple separate networks, the hardware elements of the network are, in general, utilized more efficiently. The exact magnitude of these savings will vary greatly from carrier to carrier, depending on their overall size and the number and kind of legacy networks they are supporting. The added cost of gateway equipment to convert all traffic to IP must also be taken into account when calculating net savings.
The magnitude of hardware savings from network consolidation (or whether there are, in fact, savings at all) also depends on some effects that are quite familiar to students of engineering economics but that may not be obvious to more casual observers. One is the principle of sunk cost. If you have already made the capital investment in an older network technology, simply changing it out for a newer, more efficient technology may well result in an increase in overall capital cost, because the original expenditure is sunk and can't be made to go away. This concept underlines the crucial importance of operational cost reductions with new technology, because if these cost reductions are large enough, they can pay for any near-term increase in hardware capital costs. Sometimes there is another way out of the dilemma if the older equipment can be resold or reused to displace other investments, or if it is simply near the end of its book life and tax life.
Voice compression is possible with circuit-switched networks and has been done selectively (for example, on expensive intercontinental trunks) for many years. Today, it is widely applied in circuit-switched wireless networks. However, voice compression has not been broadly used in wireline circuit-switched public voice networks. Carriers have shown reluctance to invest in compression since it affects a relatively small part of the overall cost picture; and, on the downside, they are concerned about the impact on voice quality. Voice on IP may be an enabler for voice compression in the PSTN, since VoIP systems typically come with a choice of codecs, some of which offer compression at various levels.
Relatively conservative approaches with minimum impact on voice quality include the 2 compression available with adaptive differential pulse code modulation (G. 726) and the additional 2 or so (for a combined 4 effect) from silence suppression. If one is willing to tolerate some degradation effects, standards are available for coders with bit rates of 8 kbps or even less (compared with uncompressed digital telephony at 64 kbps). The actual savings, compared with circuit-switched voice, are reduced somewhat due to protocol overhead; the exact values depend on which options are chosen for things like header compression and carrying IP packets over ATM, and unfortunate choices of these parameters can actually result in negative savings (i. e., increased costs). Within the portion of the network where compression is enabled, savings can be realized both in the transmission systems that interconnect switches and in the costs of switch ports themselves.
However, there are additional complications in the public network environment because of its multicarrier nature. Due to the regulatory situation in the United States, most U. S. long-distance calls cross two network boundaries. With deregulation, this kind of scenario is becoming increasingly common in other countries. Also, calls between countries usually cross at least one network boundary. At the present time, carriers planning for packet-based compressed voice generally assume that voice will be decompressed and returned to circuit format at the network boundary. This extra step, of course, adds to cost. It also creates the danger of tandem encoding, in which the same traffic is compressed more than once, possibly leading to a significant quality degradation and an increased delay of the voice signal. Eventually, we may see agreements between carriers to hand off voice in compressed form, which will increase the opportunities for savings and avoid the tandem encoding problem. This handoff scenario will require stable, detailed standards, not only for voice compression itself, but also for signaling and associated network management functions. It will also require good, mutually advantageous business relationships between the involved carriers. These differing scenarios for voice compression are shown in Figures 12.3 through 12.5.
Switch port costs are an interesting category. Apart from compression effects, one can try to compare the costs of terminations on a circuit switch and a packet switch that are doing something equivalent, such as handling the same number of bits per second. The packet switch usually wins, and prospects are for the gap to widen, since circuit switching is the older technology and thus is further along on its experience curve. From a fundamental point of view, it is not quite clear why this should be so. Why should the hardware be so sensitive to whether it is handling, at the same bit rate, the fixed-length groups of bits typically seen in the internals of a time-division circuit switch, the fixed-length groups of bits plus small headers of ATM switches, or the variable-length groups of bits (packets with headers) of an IP router? In part, the answer may simply be that the enterprise market where packet switching has grown up involves more intense price competition than the market for public carrier equipment, so the vendors have been forced to build it cheaper.
The most significant switch hardware cost difference, however, appears to stem from the fact that higher port speeds are actually available on packet switches than on traditional telephony circuit switches. This reduces the cost per bit per second at the switch interface, and also reduces the need for intermediate multiplexing equipment to get up to the speed of high-capacity optical transmission systems.
Another aspect of switching cost is software cost. The costs of software are either recovered through separate charges or included in port prices by switch manufacturers. Circuit switches for public voice networks typically come bundled with a large set of voice feature software. The cost of creating this software is related not only to the functionality of the features themselves, but also to the high standards for quality and ease of use that prevail in the public network industry. Thus, considerable effort is expended not only in testing and integrating features so that they do not have undesirable interactions with each other but function in precisely the same way from one release of the product to another. More than one observer has noted that expectations are very different in the desktop computing industry. This can be amusing to think about, but it really does point to a kind of cultural divide between the traditional telecommunications and information technology industries, which, in turn, is manifested in things like the relative cost of switch software.
So, what is the bottom line? Moving from qualitative discussion to actual dollar figures requires engineering and costing out some specific network configurations. W. R. Byrne of Lucent Technologies Bell Labs kindly shared some data with the authors from a recent study comparing circuit-switched with packet-switched infrastructures for new and incumbent U. S. interex-change carriers. (The packet technology in the study was actually ATM, but to the first order we can expect similar results for IP.) Results were that a new entrant using packet technology experienced a 15 percent reduction in depreciation and operations costs (about a 3 percent reduction in total costs, including access charges and SG& A) compared with an incumbent carrier using circuit switching. Savings for the packet-based new entrant came primarily from the following:
- Use of fewer but larger switches, resulting in operations support savings
- Additional operations savings from replacement of trunk group administration with link-based bandwidth management
- Small transport savings due to better packing of voice and data on one network, and higher interface speeds
To summarize this survey of interexchange carrier costs and their implications for the future of the integration of the Internet and telecommunications: The best prospects for actual cost savings appear to be in the area of operations costs for consolidated networks, and if IP-based services continue on their meteoric trajectory, this may provide the carriers with a path to get there. However, there will be many uncertainties and challenges along the road.
By the way, the preceding discussion was based mostly on the situation of U. S. interexchange carriers, including figures from AT& T, which is certainly a prototypical U. S. interexchange carrier. What about other carriers? To get some feeling for the cost components seen by different types of carriers, let's zero in on annual capital-related spending as a measure of the proportion of costs related to network hardware (as opposed to other items such as sales). Continuing to rely on financial statements from annual reports, some interesting figures are shown in Table 12.2.
Regarding Table 12.2, note first of all that we have now switched to percent of revenue, as opposed to percent of operating cost, since revenue appears to be reported on a more uniform basis across the spectrum of companies listed here. Second, while most of the U. S. companies plus Deutsche Telekom publish a figure for depreciation and amortization, others offer different indexes of capital-related spending, as described in the table notes. Finally, the authors are aware that accounting standards differ around the world, but we are very definitely not experts on that subject.
Table 12.2 Depreciation (or Other Measure of Capital Spending) as Percent of Revenue
Source: Derived from company annual reports.
- a Purchase of tangible fixed assets.
- b Amortissements des immobilisations corporelles et incorporelles.
- c Pro forma--as if MCI and WorldCom had already merged in 1997.
- d Net cash used by investing activities.
- e Cash used for investing activities.
With all of this information, it seems possible to make some sense out of the percentages displayed in Table 12.2 and to relate them to the differing network businesses run by these companies. As expected, the AT& T of 1997, as an almost pure U. S. interexchange carrier, has the lowest ratio of capital cost to revenue, reflecting the relative dominance of other costs, such as sales and advertising, in that sector. The U. S. RBOCs (Bell Atlantic, SBC, US West) and European exmonopoly national operators (BT, Deutsche Telekom, France Telecom) show a markedly higher ratio of capital spending to revenue, on the order of 20 percent. This higher rate reflects the fact that these networks own more hardware per customer endpoint, notably the local loop itself and the associated local (class 5) switch termination. To some degree, the figure may also reflect lower sales and advertising costs in these markets, which, while they are no longer legal monopolies, do not yet have the full-blown competitive dynamics of the U. S. interexchange carrier industry. Deutsche Telekom and US West come out at the high end, which may reflect continuing infrastructure building in eastern Germany in the former case and the rural sparseness of the mountain West in the latter. The MCI WorldCom figures, which by an accountant's magic make it appear that these two companies had already merged in 1997, represent an intermediate case, with the former MCI being a nearly pure interexchange carrier, while WorldCom included some significant competitive local exchange carrier (CLEC) assets. Sprint, a company with both a long-distance and a local side, winds up looking more like a local carrier on this scale.
PSInet is a very interesting case. This company is a large, facilities-based ISP. Since it is engaged in an intensive program of building and acquiring facilities, depreciation and amortization come out at the high end--at 24.4 percent of revenue, the figure is higher than for any of the telephone-type companies except Deutsche Telekom. PSInet's annual report notes that this figure is expected to go even higher as it reduces its dependence on leased facilities from other carriers and builds/ acquires more of its own.
At first glance, it would appear that opportunities for savings through the application of new network technology would be even greater in the case of the RBOCs and ex-national monopolies than in the U. S. interexchange carrier case that we examined in some detail, simply because network hardware costs are relatively greater. However, keep in mind that the increased network costs are almost entirely on the access side; so, in actuality, there is relatively less opportunity to save by consolidating traffic from different types of customers on a converged core network--the core network is a small thing compared with all the last-mile hardware going out to homes and businesses. (This may stretch the point a bit in the European national carrier case, but not for the RBOCs, which, in spite of the Telecom Act of 1995, still have very small long distance businesses.)
For these access-intensive businesses, access consolidation, rather than central network consolidation, may present the largest opportunities. Transmission technology solutions, such as running high-speed fiber access to business locations and utilizing DSL solutions to replace multiple phone lines for consumers, will achieve the first-order savings in the access area. The integration of the Internet and telecommunications may play an enabling role in access consolidation, however, by providing a common format for the carriage of multiple media on these high-speed access technologies.
So, returning to the question at hand: Is the technology driving the integration of the Internet and telecommunications relevant to the public carrier industry? All public carriers of any size are themselves sorting through the issues we have discussed. Currently, most are sufficiently convinced of the technology's potential relevance to conduct experiments or trials, and to include it in their target architectures and migration plans. In addition, a small number of new carriers have placed much bigger bets by making Internet-based infrastructures their major market differentiator.
Turning from the carrier world to the world of the enterprise, a number of things are markedly different. To see why, let's try thinking about costs in terms of the categories used in the discussion of interexchange carriers.
- Access and other interconnection. This is a set of costs peculiar to the regulatory situation of IECs and does not apply to enterprise networks in this form at all.
- Selling, general, and administrative. In contrast to the huge sums spent by IECs on sales and marketing, most managers of enterprise networks should presumably be able to get away with modest internal information campaigns to make users aware of the network and its capabilities.
- Network and other communication services. Remember that this is where the operations costs are. These will certainly be a big factor in any decision to adopt a strategy for integrating the Internet and telecommunications in the enterprise, just as they are for the IECs. For enterprises, another recurring expense associated with the network is the cost of services purchased from carriers--let's think of these costs as residing in this category.
- Depreciation and amortization. Depending on the amount of network equipment they own outright, enterprises may have substantial hardware-related costs in this category. In many cases, they will own things like PBXs, routers, LAN switches, inside wiring, and telephones.
So, to summarize, in terms of the cost categories, compared with carriers, enterprises will typically see a much higher percentage of their costs in the categories actually related to building and operating the network, as opposed to persuading users to use it or dealing with regulatory arcana such as carrier-to- carrier access charges.
These cost differences begin to hint at the really fundamental advantage that enterprises have over carriers in seeking to integrate the Internet and telecommunications and the reason why, in many instances, they will be earlier to adopt. The fact is that enterprises overall will typically have more detailed knowledge of and better control over demand, usage patterns, traffic levels, QoS requirements, and network utilization. Okay, we hear the rueful laughter from network managers convinced that they are at the mercy of an arbitrary and capricious user population. Just consider, by contrast, the world of the large public carriers. First, they must run elaborate television advertising campaigns and mount costly promotions to attract consumers to their services, and at the same time, they must support sophisticated marketing efforts to woo business customers. Then they must live in terror that some falloff in quality or lapse in network reliability will cause mass desertion to the equally heavily advertised services and promotions of one or more rivals. Finally, to engineer their networks with the proper amount of equipment to carry the traffic load at the expected high quality of service, they must rely entirely on statistical estimation methods--and face the consequences when the tried-and-true ones are thrown off by something like the sudden popularity of hour-long Web browsing sessions.
So, if you as an enterprise network manager do have, relatively speaking, a solid and detailed knowledge of your firm's network demand, usage patterns, traffic levels, QoS requirements, and network utilization, you can apply this to understand in which portions of the network the integration of the Internet and telecommunications may have immediate, quantifiable payoffs. If the transcontinental links in the IP intranet are underutilized, fill them up with VoIP. If the firm is opening a new office in Orlando, study the possible life cycle cost advantages of an IP PBX and a single wire to the desktop for voice and data. Think twice about routing the chairperson's phone calls through a voice-on-packet compression box, but think hard about applying the same technology to routine international phone traffic.
Quite possibly, you will find that the conclusion stated in this chapter's title, "The Technology Is Relevant," will apply right now to at least some part of the enterprise network for which you are responsible.
Private Networks and Carrier Services
So far, we have been discussing the economics of carriers and enterprise networks as if they were independent. They are not, of course. In fact, anyone who has been managing enterprise networks for some time is well aware that tariffs and services that are available from public carriers can have a dramatic impact on the optimum architecture choices for private networks. The rapid evolution of private voice and data network architectures through the 1970s and 1980s affords some good examples.
At the beginning of this period, it was typical for a company's private voice network to be constructed of PBXs interconnected by private analog tie lines leased from carriers, and for the data network to be based on an entirely separate collection of analog leased lines that interconnected things like front-end processors and terminal cluster controllers. The next step in the evolution occurred with the availability of tariffed T1 services from the carriers. These stimulated the development of multiplexers, which could place both voice and data traffic on the same T1 line. With the ready availability of leased T1 lines and multiplexers, it became common to design enterprise networks that integrated voice and data transport on the leased lines.
The next moves were made by the carriers. They applied the new technology of the Intelligent Network to create virtual private voice network services within the public network infrastructure, and also began to raise the prices of private lines, including T1s, to a level that better reflected the operations costs of these special services. Ultimately, the economics of the virtual private network carrier services became so compelling in comparison with private lines that many companies migrated to these services for their voice needs. Initially, most of the data traffic remained on private lines, but then a new generation of carrier data services emerged, notably ones based on frame relay technology, and significant migration of enterprise data traffic to these carrier services occurred.
It will be quite interesting to see what carrier service options for enterprise networks emerge in the era of Internet-telecommunications integration. Since we have seen that the application of packet technology to the integration of voice and data provides at least some modest opportunities for cost savings to carriers, it would be logical for carriers to develop attractively priced IP-based virtual private integrated voice-data services and to offer them to enterprise customers, effectively passing on the savings. Of course, some enterprises may be in a position to realize even greater savings with a go-it-alone approach, but for others, as in the past, carrier services may provide an attractive option.
Regulation and Its Impact
Let's turn to another topic that has generated a vast amount of literature in the academic and trade press--government regulation and its impact on the feasibility and attractiveness of network convergence.
Most writers assume that in the long run there will be relatively little regulation of services such as Internet telephony. This assumption is certainly consistent with the trend of the last two decades toward greatly reduced regulation of telecommunications in general, and, barring some unpredictable political sea change, it will almost certainly be the case. In the nearer term, however, regulation is definitely having an impact, especially on the carrier side. Some regulation benefits the integration of the Internet and telecommunications and some makes it more difficult; some regulatory effects are deliberate and others are inadvertent, caused by the collision of new technology with regulations designed with older technology in mind. Some of the strangest regulatory effects are seen in the field of international communications, where they are driven by the different rates at which deregulation is proceeding in different countries.
At one extreme end of the spectrum, quite a few countries have actually made Internet telephony illegal. This is particularly true for so-called class 1 (phone-to-phone) and class 2 (computer-to-phone) Internet telephony, which competes directly with embedded telephony providers. In some countries, the embedded telephony provider is still a government owned and operated entity that has a legal monopoly on telephone service. It is inherently difficult for a government to enforce a prohibition against class 3 (computer-to-computer) Internet telephony, because unless you look inside the packet payloads, it looks like a data application. Making it illegal to operate gateways to support class 1 or class 2 service is a little easier. Singapore, whose embedded operator will have a legal monopoly through the year 2000, takes the pragmatic approach of stating that a user who wants to engage in Internet telephony from a standard telephone set may do so only by using a gateway that is located at that user's premises.
In the United States and Europe, regulators have appeared reluctant to write too many new regulations that deal explicitly with Internet telephony. The European Commission has studied the matter and has decided that, for the time being, no action is needed because Internet telephony service is not comparable with the ubiquitous, high-quality voice service that is available over the public telephone networks of Europe.
The U. S. Federal Communications Commission (FCC) has also devoted extensive study to Internet telephony but has balked at several opportunities to impose regulation, giving the overall impression that it wishes to encourage this new form of telecommunications for its possible competitive benefits. This has left most providers of Internet telephony in the United States in the beneficial status of being information service providers that do not need to pay carrier access charges to the LECs or contribute to universal service funds. Furthermore, their rates are not regulated by the federal or state governments. So, overall, this situation is quite favorable to the Internet telephony providers, and it is not expected to change at least until, through improvements in the quality of service and ownership of their own transmission facilities, they might begin to resemble existing telephony providers so much that legal logic would require them to be subject to the same rules.
One more aspect of the U. S. regulatory environment that benefits Internet service providers in general is worth mentioning, because it may also help their entry into the field of Internet telephony. This is the predominance in the United States of flat-rate local calling. In most areas of the United States, it is possible for consumers to get a local calling tariff at a flat rate per month, independent of the actual minutes of use. This practice started long ago, when primitive electromechanical switching systems without central control units were not able to do much in the way of usage recording. When electronic switching systems were introduced starting in the 1960s, the telephone companies attempted to move to a local measured-service system, but met great resistance from consumers and public utility commissions, and largely backed off.
The flat-rate system has obvious benefits for consumers who want to dial up to an Internet service provider and then engage in Web browsing. It also helps Internet telephony by, first, creating a larger market base of ISP subscribers and, second, limiting the amount of telephone network charges incurred to enter the Internet. Flat-rate local calling is less available outside of the United States, and this has somewhat hampered the growth of consumer Internet access in other parts of the world.
One can view the availability of the flat-rate system to ISPs and their customers as somewhat unfair, since Internet users typically exceed by a large amount the average monthly usage on which the tariff is based, and thus have the cost of their long-holding-time calls subsidized by other users. When Internet usage by consumers first took off, some telephone companies used this argument along with the switch and network congestion caused by unexpectedly long holding times in petitioning to regulators that something must be done about the way ISP services are regulated. However, the telephone companies are increasingly using technology to identify calls to ISPs and route them off to data networking equipment, which is less sensitive to call holding time.
It is probably in international communications that Internet telephony has benefited most from the existing regulatory regime--one that came into being long before this technology existed. The subject of how international telephony prices are determined based on bilateral negotiation of accounting rates is a fascinating and rather complex one. The bottom line is that there are many pairs of countries for which the price of a telephone call as charged to the user appears to be far above the marginal cost of providing the call.
The international accounting rate system would appear to be unsustainable in the long run, given the probable future of competitive, privatized international networks. In the meantime, the system has been under attack from multiple directions. On the regulatory front, the U. S. FCC has initiated various unilateral actions designed to bring prices more into line with costs. In the business realm, the artificial prices create opportunities for arbitrage. One way of exploiting this, for example, is to find situations where there are very high calling rates between country A and country B, but much more reasonable rates between both A and C and B and C. A business can then be created of routing calls from A to B via C. There are also many situations where it costs much more to call from A to B than to call in the opposite direction (B to A). These cost differences have given rise to the well-known callback industry.
A different approach to getting around (or exploiting) the accounting rate system is technological--finding some way of routing international phone calls without using the international public network. This is where Internet telephony comes into the picture. If a call can be sent over the Internet from a computer in country A to a computer in country B, or, by using gateways, from the domestic phone network in A, then over the Internet, then to the domestic network in B, the international phone network and its potentially high prices can be avoided entirely.
While this Internet bypass of the international phone network appears to be a golden opportunity for Internet telephony, at least until the international accounting rate system finally collapses, there are several things to keep in mind. Perhaps the most important is that there are other, simpler technological end runs that can be made around the international public network. The most straightforward is to lease an international private line and use it to route calls that bypass the international switched network. The line can connect private networks within the two countries, and thus be used by a business for its internal calls; or the line can be connected to the domestic network on one or both sides and used to provide a cut-rate international calling service to subscribers. This is called international resale, and its use predates Internet telephony.
International resale of private lines actually has a number of advantages over Internet telephony. It can be accomplished with tried-and-true voice equipment, without the need for packet-circuit gateways. Voice quality can readily be made equivalent to that of the international public telephone network. On the other hand, there are a few advantages on the Internet telephony side. As discussed earlier, voice compression, while possible in circuit-switched networks, may be more readily available with packet voice, and, given the inherent high costs of international calling, more users may be willing to make a trade-off for lower cost, even if it involves some degradation of quality. Also, it is probably true that it is harder to regulate the use of the Internet for bypass of the international public network. This point is significant because the resale of international private lines for voice is actually only legal among a relatively small number of technologically advanced countries with progressive regulatory regimes.
So, we have seen that in the short run, regulation cuts both ways--sometimes favoring the integration of the Internet and telecommunications and sometimes making this integration more difficult to achieve without running afoul of the law. In the long run, the consensus is that all types of networks will compete on their technical and economic merits, and that regulators will gracefully claim victory for the benefits that this competition will bring to users.
Multimedia: Voice as One Application among Many
In the preceding sections, we have posited that there are at least some sets of circumstances in which the technology used to integrate the Internet and telecommunications is relevant in its simplest form--that of Internet telephony, in which the Internet plus appropriate hardware and software mimics the functions of public or private circuit-switched networks in carrying point-to- point voice calls.
A much more powerful case for the relevance of the technology can be made for multimedia applications. With multimedia, voice is just one form of communication among several that must be handled at once. Consider, for example, the case of multimedia conferencing, described in detail in Chapter 9. When one service must deal not only with voice, but also with video, text, graphics, and still pictures, a flexible, common transport medium is advantageous. As we also mentioned, the control of complex communications scenarios such as multimedia conferences is also greatly aided by the application of Internet technology, such as the Web and browser interfaces.
Also, we have been speaking about voice as a single application with, implicitly, one set of requirements. This is a habit carried over from traditional telephony, which defined voice as a 4-kHz signal and then built a massive worldwide network optimized for the transport of that signal. However, with the flexibility afforded by packet-based voice transport, different kinds of voice and audio applications can be accommodated by changing bit rates, coders, and quality options. Examples would be high-quality voice for conference or lecture applications and CD-quality audio for music. Another assumption of traditional telephony that need not hold in the multimedia environment is that voice (or audio) must be transported as a real-time signal. With packet voice/ audio, files could be sent using low bandwidth at less-than-real-time rates for later listening, or large stored voice files could be sent between voice-mail systems using high bandwidth at greater-than-real-time rates.
In considering these kinds of scenarios, for perspective, one must keep in mind that there is a kind of inverse relation between the size and complexity of the type of multimedia communications and how often it occurs. For example, in today's telephone networks, by far the greatest number of minutes of use are generated by telephone calls between exactly two people; three-way calls generate considerably fewer minutes; calls involving many endpoints and the exchange of fax and images in addition to voice generate still fewer total minutes; and so on. By making multimedia communication more convenient and more cost effective, the integration of the Internet and telecommunications has the potential to increase its use, but an inverse relationship of this kind will, most likely, still apply.
A case that is less complex than a full-blown multimedia, multipoint conference, but that may be very important because of its frequency of occurrence, is the kind of communication that often takes place today from the business desktop, in which a phone conversation is augmented by reference to simultaneously viewed Web pages, graphics files, or documents. This scenario is amenable to improvement by applying the integration of the Internet and telecommunications in any of its several forms: either by carrying all of the communication over an IP network or, in the fashion of click-to-dial and Web-based conference calling, by using the Internet and a browser interface to control the voice portion of the communication, which may still be carried over the circuit-switched telecommunications network.
Specialized Applications and Their Requirements
This section is similar to the preceding one in that it examines the relevance of the integration of the Internet and telecommunications from an application point of view. However, in a sense, it looks at the opposite case: Rather than considering multimedia applications in which multiple media are active at the same time and there may be distinct advantages at the application level from the carriage of all media types over one network, here we consider specialized applications--monomedia ones, if you will.
Of course, this is a setup. If you want to really annoy the advocate of an integrating technology (let's pick as an example ATM, which is somewhat neutral because it is not directly the subject of this book), all you need to do is describe a particular application in detail (pick one--voice telephony will do, or high-definition television, and so on) and go on to show how much more economically and with what greater quality of service it can be handled by a specialized network, perhaps one that is already in existence. The advocate of the integrating technology will reply, quite correctly, that if only all of the applications could be carried on the new, integrated network, then abundant benefits of economies of scale and operational cost consolidation would result. How do you get to that state, though, if each application viewed by itself would not experience compelling benefits from rolling over to the integrating technology?
This is a quite fundamental dilemma in the field of communications network architecture, and it is not easily solved. For the particular case of integration of the Internet and telecommunications, there are several things you could say in answer to it. As we discussed in the earlier sections on economics, it may be that the spectacular growth of IP-based applications, if it continues for some time, will make the jump to an operationally less costly integrated network easier. This would be so because, with IP traffic dominant, most of the investment that carriers would be making in an integrated, IP-based network would be needed anyway simply to keep up with the demand for IP-based services.
Also, recall that the integration of the Internet and telecommunications as we have defined it includes scenarios in which, for instance, voice can continue to ride on a circuit-switched network while integration is achieved at the control level. That may be a helpful case to cater to the needs of specialized voice (or other) applications that require the characteristics of a real circuit-switched network.
However, in the absence of some government mandate that all applications shall ride on one national information infrastructure, there is simply nothing to stop entrepreneurs from offering specialized networks designed to optimally support the needs of specific applications, whether these are based on an old technology like circuit switching or on some technology that has not yet been invented. The degree to which this happens will depend on the potential market size for such applications, the willingness of users to pay for optimized performance, and the ingenuity of inventors.
Summary and Conclusions
Here, the authors would first like to acknowledge that this chapter was written at the strict insistence of our editors, who, though strong advocates for the Internet themselves, also understand its technical underpinnings in an extremely detailed way, and so have a strong aversion to the kind of ill-informed hype that often passes for expert prophecy in the literature of network convergence. So we were put to the task of justifying, in a concluding chapter, the relevance of technologies used to integrate the Internet and telecommunications--about which we are admitted enthusiasts. We hope that we have succeeded in a balanced and restrained way in convincing you that the technology is relevant to at least a subset of the problems faced by those who have to build and operate networks on behalf of end users.
A summary of the main points would be as follows:
- The biggest long-term payoff of the integration of the Internet and telecommunications for carriers is likely to be operations savings from network consolidation. In the nearer term, most operators will feel their way along with experiments and trials, while some start-ups will deploy converged networks from day one as a market differentiator.
- Enterprises are generally in a better position to assess and realize near-term benefits from the integration of the Internet and telecommunications in selected portions of their operations due to their different cost structure and generally more controlled environment.
- The existing regulatory environment cuts both ways--sometimes it makes the integration of the Internet and telecommunications attractive where it otherwise would not be, and sometimes it makes this integration harder to achieve. In the long term, regulation will be much less of a factor.
- Some applications--multimedia, for example--are a natural fit for the integration of the Internet and telecommunications. There will always be specialized applications that may be better served by specialized networks.
A few final words: The business environment of the communications industry is changing with a rapidity that has never been experienced before. The technical landscape is also characterized by change, much of it coming about as the technical experts on telecommunications and data networking encounter each other with increasing frequency and as each community comes to better appreciate, learn about, and draw upon the vast fund of know-how accumulated by the other over decades of largely separate development. It's a pretty wild ride for those of us whose job it is to navigate the resulting crosscurrents; and it's a time of new opportunities for everyone who uses telecommunications and the Internet. Let's all make the most of it!
Meet the Author
IGOR FAYNBERG, LAWRENCE GABUZDA, and HUI-LAN LU all work at Bell Labs. They are responsible for strategizing the integration of the Internet with the PSTN. Faynberg chairs the IETF pint Working Group.Gabuzda serves on the boards of directors of the MSF and TINA-C. Lu is an editor in pint and spirits working groups and their liaison to ITU-T, where she also chairs the question on Future Network Architectures. LYMAN CHAPIN-Chief Scientist at BBN Technologies, CTO for GTE Technology Organization, and founding trustee of the Internet SocietySCOTT BRADNER-Senior Consultant for Harvard University, Transport Area Director of IETF, trustee of the Internet Society, and ISOC VP of Standards VINTON CERF-Senior Vice President for Internet Architecture and Technology at MCI WorldCom, founding President of the Internet Society, and coinventor of TCP/IP
Most Helpful Customer Reviews
See all customer reviews