January 1999

Internet Telephony Developers:
Time to Answer the Call

BY DOMINIQUE LINDEN

By now, most of us have heard the arguments in favor of voice and fax over IP. Internet telephony promises 14 times the efficiency of traditional voice transmissions, and can digitize, compress, and prioritize voice, filling up unused bandwidth to stuff multiple conversations over the same line that used to carry only one. The fundamental economics of sending voice and data over a single network have transformed voice over IP from an interesting technology into a proven business.

So far, Internet telephony is deployed mainly in the enterprise environment and by service providers offering cheap phone calls. It works, but not for everybody. Big companies and major telcos are still waiting for improved Quality of Service (QoS), security, and manageability of integrated voice and data networks.

Still, now that developers have made Internet telephony work, the next challenge is to make it work well. More importantly, developers need to understand the end user's point of view and build smart products using today's Internet telephony technology.

CALLING THE MARKETPLACE
In the enterprise environment, the first customers to deploy VoIP gateways were small and medium-sized companies with international sites. According to Probe Research Vice President Hilary Mine, most of the 2,000 enterprise VoIP sales in 1997, as well as the 3,000 sold in the first half of 1998, were to those companies. Enterprise VoIP sales are also expected to accelerate among other small to mid-sized companies where voice costs are a major expense, especially those already experienced with call centers or computer telephony. These companies often have an abundance of bandwidth because over-provisioning is the easiest way to avoid end user complaints about throughput delays and, thus, about carriers' private line pricing policies.

Big companies are already running VoIP trials and pilot projects, but holding off on major deployments until they are convinced that QoS mechanisms can keep their voice and data traffic from damaging one another on shared IP facilities. Reliability features are not critical for super-cheap VoIP calling services, but if providers want customers to depend on their VoIP services the way they depend on the PSTN today, they must provision these services from more robust and reliable platforms. To the victor go the spoils, and developers who figure out how to make voice and fax over IP work for large enterprises and service providers with mission-critical applications will be the ones who succeed. The objective is not necessarily to improve the fundamentals of IP telephony - it's enough to build a solution that works today. There is no time to wait for better products. It took the telephone 38 years and the fax machine 22 years to reach 10 million users. The personal computer reached 10 million users in only nine years and the Internet took less than two. Time is running out.

THE MYTHS OF INTERNET TELEPHONY
There are still many myths about Internet telephony, including voice quality problems because of packet loss, delay, and a lack of sophisticated voice codecs. None of these are true. Voice quality has improved dramatically and should no longer be of concern to developers. Today's Internet telephony systems deliver voice quality in a box. In the early days of geek-to-geek communications, IP packets could get lost. But those days are gone. Intelligent gateways provide seamless voice and fax communications, and if lost packets were really an issue with fax transmissions, you would see blank lines on the page. Unlike our mind, a machine can not compensate for words that were not transmitted correctly, and lost packets in fax over IP would show black on white. As the first real-time fax over IP applications come to market, it is obvious that lost packets are a myth of the past.

Another misconception is that the plain old telephone is the ultimate standard for voice quality. Wrong again. The 64 Kbps compression mechanism in the phone network was chosen randomly. Now we don't need 64 Kbps to achieve the same kind of voice quality. A whole new generation of high bit-rate coders (higher than 64 Kbps) are delivering better sound quality than the regular telephone with today's increased bandwidth availability. In a few years, everybody will demand CD-quality digital sound, and people will complain about the poor performance of the plain old telephone.

SO, WHAT'S THE PROBLEM?
If voice coders and packet loss are not causing voice quality problems, then what are the big companies waiting for? First, delay is a function of the gateway and the network, and people think they have to wait for better gateways before deploying Internet telephony. In reality, they must redesign their network. There is an increasing need for people who understand the whole system that makes up an IP voice and fax solution, and who understand what QoS is all about. QoS has nothing to do with voice quality - it indicates how well the system functions as a whole. What matters is how good the service is for the end user, who will judge it based on voice quality, reliability, and manageability. Therefore, most gateways already offer a variety of functions on Windows NT and UNIX servers, routers, and remote access servers with a healthy selection of interfaces, compression algorithms, and voice quality features like echo cancellation and jitter buffer.

Today's Internet telephony systems lack the system management tools to improve QoS. There are no easy-to-use configuration and management tools for VoIP networks, no call detail records (CDRs) for charge-back to customer departments, and no reasonable security measures. Gateways do not support true redundancy or hot-swappable components, common features of data switches and routers, and few have an SNMP agent onboard. This can be problematic because end users of voice and fax over IP expect at least the same quality and features they get from the PSTN network. But voice over IP solutions are implemented by IT managers with only one thing on their minds: keeping their jobs. An IT manager's job is not to be a pioneer, but to make sure that calls go through, no matter how the network is configured. Both can be accomplished by focusing on QoS instead of only on voice quality.

MANAGEMENT AND SECURITY SOLUTIONS
Network managers insist that any device connected to their networks be manageable via SNMP, part of a more structural approach to QoS. Internet telephony platforms can be very bandwidth-smart by fully supporting the RSVP and IP precedence specifications, which allow for traffic prioritization and optimization. RSVP and IP precedence are features of the router, not the gateway. The gateway, however, must be able to flag voice packets so that they will be given priority over others when arriving in the router. IP Precedence and RSVP are becoming the two most useful signaling mechanisms because both take advantage of the end-to-end nature of Layer 3 protocol and the growing ubiquity of IP as the network protocol of choice. IP precedence uses the three precedence bits in the Ipv4 header's QoS field to specify calls of service for each packet. Managers can partition traffic in up to six classes of service using IP precedence.

So far, security seems to have taken a back seat. But before Internet telephony becomes mainstream, airtight security for Internet telephony connections needs to be addressed. Authentication (are you who you say you are?), integrity (did anybody tamper with the data?), privacy (did anybody tap into a conversation?), and non-repudiation (proof that you really made a phone call) are all addressed through H.235, formerly known as H.Secure. H.235 is a new member of the International Telecommunications Union's (ITU) H.323 standards umbrella for setting up VoIP calls. Firewalls should also be implemented with H.323 proxies to allow H.323 calls through the firewall without opening it up to all kinds of traffic.

DEVELOP FOR YOUR USERS
Having said all this, what's more important than QoS management techniques is taking the end user's point of view when designing an Internet telephony solution. A good example is choice of operating systems. Some developers think it's impossible to design reliable and redundant systems based on Windows NT, so they develop on UNIX and are waiting for CompactPCI hot-swappable hardware. But do end users really care which operating system is used? No. All that matters is that the phone service doesn't go down. Some developers believe in Windows NT because it's easy to program, and toolkits and management tools are readily available. They, too, achieve redundant systems. It may not be graceful from an engineering point of view, but in an emergency you can hot-swap an entire configured server instead of taking out elements like voice boards. After all, we are living in the throwaway society of complete product replacement.

What the IP telephony industry needs most is developers who can think like end users. Instead of focusing on the technical difficulties of Internet telephony, developers should work on end user solutions. Bill Buxton, human computer interaction specialist, once said it very nicely, "Let's make smart products with the stupid technology of today, instead of making stupid products with smart technology of tomorrow." There is no industry for which this is more appropriate than for Internet telephony.

Dominique Linden is product marketing manager for Dialogic Corporation. Dialogic is the leading manufacturer of high performance, standards-based computer telephony components. Dialogic products are used in voice, fax, data, voice recognition, speech synthesis, Internet telephony, and call center management telephony applications. For more information, visit the Dialogic Web site at www.dialogic.com.