The insider's guide to IP-based telephony: development, network deployment, testing, and beyond
Now that virtually every leading telecommunications service provider has committed to delivering IP-based telephony services, communications professionals face the enormous challenge of implementation. This hands-on guide brings together today's best-known answers and solutions for delivering Voice Over IP (VoIP) services with the quality customers demand. No other book covers the combined issues of protocol signaling, media transport methodology, reference topological considerations, and voice quality testing in service offerings. Bill Douskalis presents systematic coverage of every aspect of IP-based telephony:
No matter what your role in delivering VoIP services, IP Telephony delivers the specifics you need to speed deployment, improve reliability, ensure quality, and simplify troubleshooting. Precise, thorough, and based firmly in the real world, it is simply indispensable.
1. Protocols for Voice Over IP and Interactive Applications.
IP Telephony fundamentals. SGCP, MGCP and NCS—Overviews. H.323. An Overview of the ITU-T Approach for VoIP Telephony. SIP—Summary of Features of the Session Initiation Protocol.
RTP and RTCP—Definitions for Media Transport in IP Networks. RSVP The Resource Reservation Protocol. A Quick Look at AAL2 for Voice Transport. Edge Bandwidth Management—A Look at AAL2.
An Overview of SS7. The Services of SSCP and TCAP. Call Control in Packet Networks. Heterogeneous Call Setup—Mapping SS7 to IP-Based Call Control. Closing Thoughts on Mixed Network Voice Telephony.
Voice Quality in Converging TDM and IP Networks. Auditory Perception and Testing Methodologies. Perceptual Speech Quality Measurement. Voice Quality Test Instrumentation. Echo Measurement and Control. Summary.
Objectives. From the Big Picture to the Lab. Equipment Setup and Configuration. Baseline Measurements. Testing with Connected Gateways. Observations and Conclusions.
A Look at a Real-Time Trace of the T.30 Protocol. Fax over Packet Networks. Real-Time Internet Fax—T.38. Basics of V.34 Fax. Group 3 Fax Error Correction Mode (ECM). Voiceband Data Transport Revisited—AAL2. Summary of the Fax Discussion and Possible Issues. References.
In the last five years of the twentieth century, we have seen developments in the telephone network which will affect our lives well into the new millennium. A continuing and massive effort by the industry is bringing closer the day when integrated services and multimedia will finally come to our homes-on demand and at affordable prices. The Plain Old Telephone Service, POTS as it has been known for most of the twentieth century, is giving way to modernization, service integration, and convergence of two fundamentally different types of technologies: voice and data.
There is nothing wrong with the quality of the Public Switched Telephone Network, the PSTN as we have known it for so many years, either in the delivered speech quality or in its reliability. On the contrary, it is the model of robustness and security, and those of us who know its intricate details are amazed by how something this large and this complex can function so reliably and so well. But the quality of a telephone conversation seems to have given way to other, higher priorities and is no longer the only driving factor for modernizing the network. Changing needs in both the business and residential market segments require new telephony features and capabilities that are neither simple nor economical for service providers to create on top of the PSTN infrastructure. This contrasts the current fact of life of the telephone companies, which are approaching the point of not being able to produce a single new feature in their present networks without substantial expenditures and delays. Service and feature creation on standard Class 5 service platforms became simply an intractable and very expensive proposition. So, when the requirements of the market aligned with the needs of the service providers, it was only a matter of a short time before it became obvious the PSTN had to be transformed and modernized, not in an evolutionary manner but in a revolutionary manner. But is the new PSTN that is coming our way going to be better? And who defines what better is? And who will pay for all this modernization? There are both simple and complex answers for these three questions.
In the mid-1990s when I started consulting outside my immediate home base, it became necessary to communicate electronically with my customers. ISDN BRI 128Kbps service did the trick to provide sufficient data bandwidth for web access and receiving and sending 10 Mb+ email attachments in moderate time, but ISDN service never became a commodity telephony product. In fact, part of the reason we are witnessing the modernization of the PSTN is because service providers cannot offer us enough bandwidth at a reasonable price, because it costs them too much money to add and upgrade existing network equipment. And even after they do it, we still do not have benefit of service integration and multimedia capability because 128 Kbps is simply not enough bandwidth to converge voice, video, and data services. The world's brightest minds have pursued a feasible and economical solution, and in the last couple of years major decisions have been made to steer the network revolution in specific directions.
Many papers have been written in the past few years about the need to bring integrated telephone services to the customer premise, and they are all informative but somehow the big picture has not been completely described. Part of the reason is that things are changing very rapidly-even as this book is being written. Another reason is one's perspective of what integrated telephone services will be, which may not always encompass all aspects of the new public network. It is not easy to develop a complete perspective of the public network details without getting intimately familiar with telecommunications technology, but I believe an effort to present the high level big picture is worthwhile even in simplified form. So, let's take a look at a concise view of what the PSTN is today, what is changing in the new world, and what is staying; how the changes are being approached and the reasons why the reference points have been selected for change. It would take hundreds of pages or more to go into every detail of the PSTN architecture, but I hope a simplified view will make it easier for the reader to understand the big picture.
The PSTN is the collection of all the switching and networking equipment that belongs to the carriers which are involved in providing telephone service. When we talk about the PSTN, we mean primarily the wireline telephone network and its access points to wireless networks, such as cellular, PCS, and satellite communications. Subscriber access to the PSTN wireline network is through large voice switches, which are located in a telephone company central office, and bring us basic telephone service through ordinary analog telephones or digital PBX systems. Access can be either through wireline telephones, or via a wireless network.
Voice from the telephone set in our homes to the central office exists either in analog form for basic service or in digital form if access to the carrier is via a PBX. However, once it reaches the central office, voice exists in digital form on the PSTN-on time division multiplexed channels of 64 Kbps each, which carry pulse code modulated voice samples. There is never any doubt as to the kind of speech quality we will experience when we make a call on the TDM network, and the quality of service for a voice call has never been a negotiable parameter. However, delivering this quality comes at a huge cost to the service providers who need to tie up network resources for the duration of an entire call, regardless of the actual bandwidth utilized by the conferring parties at any instant. Even simple traffic analysis using the fundamental Erlang models shows a steep linear dependence of network resources to the user population served by the network. Furthermore, the enormous appetite for web access in the last few years has caused an even bigger problem in the TDM network-network resources, which are always voice-grade TDM circuits, are tied up while we browse the web at our leisure. Clearly this is an uneconomical proposition for the telephone companies who find themselves unable to enhance their networks and provide additional services in the current public network infrastructure. The culprit has been widely recognized as the lack of integration between voice and data services and the low bandwidth of the analog links in the "last mile."
How does one go about fixing this problem? It isn't easy and it is also very expensive. Since the mid-1980s it has been viewed as necessary to replace the TDM public network with a universal and ubiquitous packet-based alternative with nodes and links whose bandwidth can be managed in a dynamic manner all the way to the edge and onto the customer premise. Part of this promise seemed to be fulfilled with the ATM technology, which is still a major key for revamping the backbone such that application convergence can be realized. But while ATM technologists spent years developing accurate and broad standards to support the new age of the public network, the Internet experienced explosive growth and brought us a taste of what convergence and integration would be like, if we could only merge technologies and get enough bandwidth to the premise. The Internet Protocol, our beloved IP, is now the king of network protocols, up to which virtually all applications seem to looking to become part of the new convergent network. The first two are voice telephony and data applications. Still, simple convergence of a voice telephone call with a data application that requires bandwidth management has proven to be a non-trivial task. The requirements are often conflicting and the proprietary early implementations of convergent networks must eventually become interoperable in the public network to gain wide consumer acceptance. At the same time, each provider is trying to get a jump on the market. These issues have been at the heart of the problem and work continues at a frantic pace.
A simple call between two parties with analog phones is not simple, even with the current scheme of things in the public network. First and foremost is the fact that there are really two "clouds" in the PSTN: the one that carries signaling and the one for voiceband transport, as shown in Figure I.1. Part of the problem with convergence in the current infrastructure is that signaling at the local access for basic service is hardware-based and tailored for voice telephony, with some fundamental considerations for point-to-point video over ISDN circuits, for those who can afford it. This is a severe limitation and a good reason to pursue change. The second aspect is the difficulty for bringing interactive multimedia applications, such as distance learning, to the home. Both the signaling and the lack of sufficient bandwidth are severe limitations. On-demand entertainment and shopping convenience are also driving factors for the pursuit of network modernization. Affordable full-color, full-motion video conferencing with whiteboard capability may not be far away in the new telephone services. All these new capabilities may have the side effect of rendering us to couch potatoes, but they do have serious benefits by making possible the exchange of information in a manner and convenience we have never experienced before.
If we start by assuming the access bandwidth issue will be resolved somehow, the remaining technical issues form a mountain of problems in their own right. Starting with the residential customer premise, signaling and media transport must be re-engineered in their entirety to facilitate convergence. The analog local loop will be a thing of the past and voice will exist in digital form end-to-end.
The packet technology of the new infrastructure is being designed on IP-based protocols and the new connectivity will be provided over various link layers, such as ATM. A single signaling technology has not been determined yet, but three front runners-MGCP, SIP, and H.323-have set the pace in all new equipment and service designs. This means we are starting out immediately with signaling incompatibilities at the packet layer among carriers, and if we can't signal, we can't make a call. This potential impasse will be avoided by the natural inertia associated with rolling out new telephone services. Since it takes enormous expense and effort to offer packet-based telephony, the first step for the service providers is to offer it in pieces. The first piece, which has been in the works for the last couple of years, offers integration of service up to the central office. This is now a re-engineered Class 5 system that offers most of the popular telephony features of the TDM network, but over IP-based technology. When this step is complete, we will be pretty much where we are now, from the perspective of telephone service features perceived by the subscriber; but now the horizon is cleared for the providers to start adding value to the basic telephone service.
Media transport will be packet based and there appears to be no controversy that the Real Time Protocol (RTP) is the method of choice for VoIP service, regardless of the signaling protocol used to set up calls. But there is a new headache coming up, which the PSTN does not have. In the effort to converge applications and manage bandwidth more efficiently, compressed voice telephony will start making inroads in residential service. This may have as yet unseen ramifications in the wider market because objective measurements on delivered voice quality only recently became feasible. In fact, measurement of voice quality is one of the major topics of this book. This issue inevitably leads to the next discussion of quality of service classes. There is a lot of work to be done in the next few years in the area of QoS in the converged network, and only time and actual data will tell us if assumptions and projections we have made about the design of the new public network are correct.
The real problem begins once we leave the Class 5 domain and start getting into the backbone network. In the PSTN, all carriers are interoperable through SS7 signaling and PCM voice encoding. Plain and simple. Replacing the core PSTN, however, with an untested and non-uniform packet network may be a bit too much of an undertaking, and this is recognized by both the service providers and equipment vendors. So, the PSTN is expected to survive for some time, and the new packet-based Class 5 domains will become gateways to the PSTN for voice telephony. This means packet-based-to-SS7 signaling interworking and media transcodings will be a necessary fact of life. This may come at the cost of some quality degradation, but will have the immediate benefit of interoperability between dissimilar packet-based VoIP technologies through the TDM network. After all, what good is any new technology if we can't make the same telephone calls we always could?
Service integration even with the PSTN in the middle, however, will still be a powerful offering. It's a matter of the new Class 5 domains also becoming gateways to the data networks and the Internet. There they will serve as funnels and aggregators, neither of which is pretty as an ultimate solution, but it will be effective in delivering the promise of convergence as a starting step. The first things we will notice are changes in the ISP dial-up paradigm, with faster and ubiquitous service access, higher mobility of our service profiles, and more fixed and mobile telephones with a richer feature set.
All this means is that the new packet-based Class 5 domains will treat everything as data-packet data-but some data packets will be more equal than others. Indeed, the continuing presence of the PSTN in the backbone may mitigate the potential problems of convergence into a single network, because for some time to come we will not have to decide on how to treat the different priorities of data through single routing points. As we learn about emerging patterns in service usage in the new network, whatever decisions are made to place a piece of the PSTN into a single IP-based backbone will be a more educated guess at the time.
There are other major issues associated with this modernization process. The first one deals with security. The PSTN is extremely secure because the signaling network is closed to the outside world. The same cannot be said of IP-based networks, but the lessons learned in the last few years have raised the consciousness level in the area of security; and protocols like IPsec are expected to play a major role. Even so, the potential of signaling and media traversing the same links as the modus operandi of the new network is bound to be met with some skepticism from many of those who put the present PSTN together.
The second potential issue deals with continuing support for applications that are working just fine in the present TDM network. The biggest one is facsimile. Facsimile represents 30-40% of corporate phone bills and is a relatively large revenue stream for the telephone companies. The desire of the business users is to lower the cost of sending a fax, while as a minimum maintaining the status quo of the current feature set and quality of the overall user experience. This business driver has already resulted in more than one method to bring IP-based fax to the customer premise for users with varying needs and budgets.
Another potential issue is that there are numerous federal regulations that govern both POTS service and the carriers. It is hard to believe there will be no changes to the regulations, but at this time it is too soon to tell the degree to which the consumers and the service providers will be affected.
Finally, let's not forget the international telephone networks with which we have to interwork at their access points to the PSTN. It will be some time before the public networks around the world become packet based, and this will mandate continuing international interworking as the new public network is modernized.
Figure I.2 shows where we are headed with telephone services as part of a convergent network, regardless of what type of carrier brings it to our premise. The short term paradigm will consist of packet-based networks connecting to the PSTN for ubiquitous telephone service. Calls between subscribers of the same carrier can be routed over either their own piece of the PSTN or their packet infrastructure without touching the PSTN. The long-term view and desire is for one packet network to connect all carriers and all services in a compatible manner, but based on the current state of the art this reality is still a few years ahead of us. But it will indeed be a reality some day.
In order for VoIP to succeed as the forerunner to the convergent network, it is imperative for the new services to preserve the user experience of the simple two-party POTS telephone call. But it has long been argued that any packet-based technology will introduce delay in the voice stream, which can at times be perceived by the user during the conversation and yield unsatisfactory results. VoIP technology in the public network must overcome this potential problem and bring delays down to non-perceivable levels. Another factor is the voice quality delivered by VoIP in conjunction with the newest voice compression technologies. If we combine the potential delay issue with inevitable reduction in quality from voice compression, the user experience may suffer as a result. Voice quality can now be measured in an objective manner, and the gap can be bridged through the availability of more data, information, and knowledge gained from actual experience.
Finally, telephone service is also measured by the degree of robustness of the provider's network. Service availability is also a parameter regulated by the FCC. The PSTN is a "five nines" network, operating with 99.999% specified availability, and in reality even more. A reduction to "four nines" or "three nines" will mean more frequent outages and consumer dissatisfaction. This is one more reason to proceed slowly into the core network and ensure all aspects of quality are met before a piece of the PSTN is renovated.
There is obviously money at the end of the rainbow to justify the enormous expenditures in equipment and labor to revamp the public network. When voice over IP first appeared on the Internet, it was exciting for the public at large for its potential of making long distance phone calls for free. There are still people who associate VoIP with free telephone service over the data networks, but the new services will not be free and some people will be disappointed to find out there is a bill for these new voice services. So, who will pay for all this renovation? The consumer is a good answer. We will indeed get the benefits of a converged network with the promise of the information superhighway at highway speeds but hopefully at an affordable price. The service providers will get the benefit of a single, unified multiservice network some day with favorable scalability to ensure their continuing economic survivability. And the overall benefit will be greater availability and flow of information, very convenient and secure electronic commerce, simplified distance learning, and other uses and applications we have not yet imagined.
This is a technical book. We will not discuss business or social issues resulting from the modernization of the public network, but we will concentrate on some of the major technical aspects of the new VoIP services for the purpose of bringing to light relevant issues and potential solutions. In the first three chapters, we will be looking at the new signaling protocols in the new public network and the enterprise and cover the details of transporting media across similar and dissimilar network domains. We will visit the SS7 signaling protocol of the PSTN in a high-level overview but with enough detail to make the point as to where changes may be coming with integrated VoIP signaling; and we look at call flows in homogeneous and heterogeneous signaling topologies per the definitions given in the text. The reader is expected to have basic familiarity with the IP protocol as well as TCP and UDP, but intimate bit-level familiarity is not necessary to understand the concepts. Throughout the text we show several IP signaling packets, captured live with the help of the Hewlett-Packard Agilent Technologies Internet Advisor.
In Chapter 4 we cover the subject of perceptual voice coding, voice quality measurement techniques, and discuss in detail the Perceptual Speech Quality Measurement (PSQM) algorithm, as implemented in the Hewlett-Packard Agilent Technologies Telegra? Voice Quality Tester (VQT). Chapter 4 also discusses network parameters which relate to voice quality and affect the overall user experience, such as echo. In Chapter 5 we move to the lab where we set up a small reference network to take speech quality measurements and discuss the issues of delay and clarity in proper context. This is the first time we see network impairments and voice compression included in the results of objective speech quality measurement. Finally, in Chapter 6 we discuss facsimile support in IP networks and close the text with a detailed discussion of V.34 fax, which seems to be the final frontier in facsimile using the analog facilities of the PSTN.
October 31, 1999