February 1999

Overlapping, Even Competing, But Ultimately Harmonious
IP Telephony Standards Proliferate, But Retain Focus On Interoperability

BY BROUGH TURNER

H.323, SGCP, IPDC, MGCP, SIP, SAP, SDP, MDGC, IETF, IMTC, TIPHON, ITU SG 16 .… We have a veritable alphabet soup of standards and organizations developing standards for IP telephony. But - even with all of these standards and organizations - how many of us can place IP telephony calls to each other?

Ultimately, it isn't the number of standards that will determine the value of IP telephony. Rather, it is the number of people who will be able to share IP telephony calls. This notion about value echoes an idea expressed by Bob Metcalfe, the inventor of Ethernet. According to Metcalfe, the value of a network increases as the square of the number of people who can use it.

Metcalfe's idea, which has become known as Metcalfe's Law of Networking, applies very broadly - far beyond traditional data networks. Indeed, it applies to any "network," that is, any group of interconnectable devices. Fax machines, for example, form a network for the purposes of Metcalfe's Law. The reason my fax machine is so valuable is that it can connect to millions of other fax machines.

If we decide that Metcalfe's Law is indeed applicable to IP telephony, we would expect the value of my IP telephony service to increase dramatically as the service reaches more people - as more and more people are able to receive my calls. So, for IP telephony to succeed, IP telephony devices must interconnect with each other and, more important, with the large base of existing telephones, which is roughly one billion, worldwide.

No single vendor is large enough to set a standard, so growth of the IP telephony industry will require agreement on interoperability standards and then deployment of equipment that actually interoperates. Luckily, the IP telephony industry has focused on interoperability from the start.

EARLY IP TELEPHONY
Commercially viable IP telephony started to surface in 1995, principally among students and hobbyists. Actually, packet telephony had taken place as early as 1973, when ARPANET users relied on the Network Voice Protocol (RFC 741) to hold conference calls.

The turning point came in 1996, however, when Intel Corporation worked out an arrangement with Microsoft. The avowed purpose of the arrangement was to facilitate video conferencing, which would admit, however, the happy coincidence of stimulating demand for powerful processors. Intel, its interests clear, endorsed the H.323 family of standards. Intel wanted Microsoft to support H.323 as well, and to include audio and video conferencing software in Windows.

Now, the H.323 standards for audio and video conferencing over packet-based LANs had already been developed by the International Telecommunication Union (ITU). So far, so convenient. There was, however, a slight complication.

Unfortunately, some patented technology is used in the H.323 standards - most notably for the G.723.1 speech compression algorithm - which means that implementers owe royalties. Although the details have never been published, it appears that Intel worked out a deal under which Microsoft would receive the intellectual property rights, with no royalties, for a single channel of G.723.1 per system. Whatever the terms, Microsoft ultimately lent its support to the H.323 specification.

H.323 - A GOOD BEGINNING
Since Intel and Microsoft decided to endorse it, H.323 has served the budding IP telephony industry fairly well. It supports multi-media communication over a network - not just audio but also video conferencing. As a written specification, it has provided a rallying point from which the technology can expand. And, by mid-1998, we saw H.323 interoperability testing where the majority of participants were actually able to interoperate!

H.323 ISSUES
But there are also problems with H.323. Because it originated in the telephony industry, its structure and style differ from those of most Internet specifications. These differences have resulted in a clash of cultures. The Internet standards process focuses on rough consensus and working code. The ITU process involves working out a complete solution framework, but then leaving the implementation details to others, typically the local public telephone administrations (PTTs). Plus, with H.323, only the lowest protocol layers leverage Internet standards. Everything else is inherited from the telephone industry. The H.323 family of standards is also huge by Internet standards. Developing H.323 software takes time. Commercial implementations are available, but they are expensive.

There are other problems with H.323, beyond its size. With the initial version of H.323, call setup is very slow. Each call requires the exchange of multiple messages. On heavily loaded circuits, the exchange of messages may take many seconds. Compare a network that complies with the initial version of H.323, which is a network on which call setup may well be delayed, with today's public telephone network (using SS7 signaling), where the caller can dial a long-distance number and, within a second or two, hear the ringing signal from the other end.

The latest revision of H.323 reduces the number of messages that must be passed, and it goes a long way towards reducing call setup delay. But, because Microsoft has not yet adopted the new revision, H.323 retains the reputation of being slow.

Another complaint has been that H.323 doesn't scale very easily. This complaint is heard from those focused on providing IP services that are paid for by the minute. Their actual complaint may be that H.323 is not well suited for charge-by-the-minute services, though even this is arguable. H.323 actually defines a "gatekeeper" function that allows a central authority to administer bandwidth for multiple H.323 terminal devices. The H.323 gatekeeper provides many of the needed functions, but it was designed so MIS directors could control bandwidth utilization on a LAN, not so Internet Telephony Service Providers (ITSPs) could bill for service.

But H.323 is real today. It is widely deployed thanks to Microsoft's NetMeeting product. Indeed, it is the only protocol for media-rich endpoints that is in widespread use today. H.323 is not going away anytime soon.

TIPHON AND MGCP - THE NEXT STEPS
While H.323 is the only widely deployed standard for IP telephony terminal devices, most of today's gateways between the PSTN and ITSPs use proprietary protocols. Several competing efforts are addressing standards for gateways between IP telephony and the existing PSTN.

A major project called TIPHON (Telecommunications and Internet Protocol Harmonization Over the Network) is proceeding within the European Telecommunications Standards Institute. The TIPHON project assumes H.323 and works to define the rest of the system needed to provide IP telephony services that interoperate with the existing PSTN.

Meanwhile, within the Internet Engineering Task Force (IETF), separate working groups are developing standards for media gateways and signaling gateways. There is obvious overlap with the TIPHON efforts, but major emerging carriers in the United States are not going to wait until they hear the results of a European telecom standards process, a process where they have only secondary influence.

The IETF working group is evolving a specification called Media Gateway Control Protocol (MGCP). MGCP is a compromise between two earlier proposals, SGCP and IPDC, each of which has been implemented by their proponents. Here you can see the advantage of the IETF's focus on "working code." Some major carriers and major emerging carriers, like Level 3 Communications, are basing their new IP telephony networks on gateways that can be upgraded with successive versions of MGCP as they become available. So, there is substantial momentum and commercial product development efforts, even though the development of the standard is ongoing. For this reason alone, it seems likely the IETF's media gateway protocol will dominate.

BEYOND MGCP
The gateway standards efforts are focused on the one commercial IP telephony application being deployed today - toll bypass. Many emerging carriers are making money at international tariff bypass using voice over IP (VoIP) technology. And large carriers - such as Qwest and Level 3 (and, potentially, ATT, MCI/Worldcom, and Sprint) - are also looking to provide VoIP services nationally and internationally. The voice market looks like a gold mine to data network providers because it is still possible to charge for voice services by the minute. But how long will that model last?

MGCP is based on concepts originally developed in the telephone industry for the advanced intelligent network (AIN) which assume that the logic for providing services is centralized. This structure supports the ability to charge for voice minutes. The Internet and other IP networks, however, do not differentiate between voice bit and other bits, and they focus on intelligence at the edges of the network.

It is now widely accepted that IP networks will replace the PSTN. But with one billion existing telephones, change to a new model will not happen overnight. MGCP will last as long as there are traditional circuit-switched telephones to connect to. Then as new telephone-like audio appliances emerge that connect directly to the net, the need for MGCP will diminish.

So, what's the long term outcome? New specifications coming out of the IETF transport group provide some clues. Their goal is to provide different grades of service over native IP. While the resulting services are sometime referred to as gold bits and brown bits, the "color" of the bits does not indicate the type of bits - such as voice or data - but rather the class of service those bits will receive across the network. Gold bits, for example, may be guaranteed lower latency, while brown bits may get best-efforts delivery with no guarantees. Gold bits will cost more than brown bits, but the rates are likely to be based on bulk monthly levels, or flat rates as long as gold bits are no more than x percent of the total bits.

As these new services - and telephone-like IP audio appliances - appear, IP-to-IP calls will no longer be billed by the minute. It's just bits, even if we pay a little extra for gold bits. The only per-minute charges, and the only time one will use MGCP, will be to access a gateway to connect to a legacy telephone. With new telephone-like IP audio appliances that plug into Ethernet, new protocols will emerge, relegating MGCP to the gateways.

An example of things to come is already visible in the family of fairly simple protocols known as SIP, SAP, and SDP. (These refer, respectively, to Session Initiation Protocol, Session Announcement Protocol, and Session Description Protocol.) SIP, SAP, and SDP can be used for voice and for multi-media applications - unlike MGCP, which tends to be focused on voice only. SIP competes with H.323, but is simpler, layered, and Internet "friendly." In addition, with SIP, software implementations are available for free.

Despite these advantages, there are few SIP deployments as yet. Look for SIP to appear in IP-based PBX environments first, and then elsewhere as IP audio devices (telephones) proliferate, and as differential services support high-quality connections without the per-minute charges of the MGCP-focused carriers.

PLANNING FOR THE FUTURE
Change will be a feature of the IP telephony industry for many years to come, and the rate of change is likely to accelerate. But despite the development of overlapping (and even competing) standards, both service providers and equipment manufacturers are focused on interoperability.

Everyone already agrees on basic media streaming using RTP. Furthermore, the IETF efforts to develop a Media Gateway Control Protocol (MGCP) assume this new protocol must interoperate with devices using H.323 and SIP standards. And even competing standards like H.323 call control and SIP can be made to interoperate using a relatively straightforward software stack running on the endpoint, the gatekeeper, or any nearby server.

So, unlike fax, where the industry languished for decades before achieving interoperability (and explosive growth!), the IP telephony industry has been focused on interoperability from the start. Despite diverse commercial and political origins, each "competing" standard envisions ways to interoperate with other standards. The signs are good and the industry potential is enormous.

Brough Turner is senior vice president of technology at Natural MicroSystems, a leading provider of hardware and software technologies for developers of high-value telecommunications solutions. For more information, call Natural MicroSystems at 508-620-9300 or visit the company's Web site at www.nmss.com. E-mail to the author ([email protected]) is also welcome.