Convergence /

The enterprise VoIP update

Survey says deployment is being slowed by too many standards, not enough interoperability and concerns over quality of service.

By Edwin Mier, Network World GLOBAL TEST ALLIANCE
Network World, 08/27/01

Voice over IP isn't making much of an impact in companies just yet, but deployments are expected to pick up considerably in the next two years, according to the results of an exclusive Miercom/Network World survey of nearly 100 voice-over-IP equipment vendors.

To get a sense of where the voice-over-IP market is and where it's _headed, Miercom received detailed responses from 96 voice-over-IP equipment vendors including 3Com, Altigen, Avaya, Cisco, Clarent, ECI Telecom, Ericsson, Lucent, Motorola, Nortel, Nuera, Shoreline, Siemens and Sphere.

When asked how much of the total U.S. voice traffic is traveling over IP, the vendors estimated that nearly 8% of all calls today travel at least part of the way on an IP network. They also estimate that the voice-over-IP installed base today is more than 70% carriers and ISPs, and less than 30% enterprise networks. That means voice over IP in the enterprise represents closer to 2% of total U.S. voice traffic.

VoIP vendors pass SIP test
Open-door policy
Conduct of survey
The changing VoIP landscape (chart)
VoIP pros and cons (chart)
SIP interoperability (chart)

Vendor free for all: See how vendors' products interoperated with others'
(a PDF file).

Vendors are painting a rosy picture of the future, however. They predict that in five years more than 36% of U.S. phone calls will be voice-over-IP-based. And most predict the imbalance between service provider deployments and enterprise deployments will begin to level out over the next two years. Fifty-five percent say they believed voice-over-IP port shipments over the next two years will be closer to a 50/50 split.

Protocol proliferation

On the issue of standards, we found that Version 2 of the ITU-T's H.323 specification still reigns as the most commonly implemented voice-over-IP call-control standard. But that is likely to change in the next year.

While few respondents say they are planning to add H.323 Version 2 support (as nearly three-fourths say they already do), more than half plan to add support for the latest version of the IETF Session Initiation Protocol (SIP) and/or the ITU-T's new H.248 voice-over-IP call-control standard (see graphic, right).

With 30% of vendors claiming to already support SIP, and 51% more planning to add SIP support, more than 81% of the industry could be supporting SIP by next summer.

Media gateway control protocol (MGCP) support, which seemed neck and neck with SIP support just a few months ago, appears to have lost some momentum. About half of the voice-over-IP vendors say they expect to be supporting MGCP by next year.

The choice of which control protocol to implement is key for different vendors' voice-over-IP gear to be able to interoperate. And this proliferation of protocols does not bode well for ubiquitous voice-over-IP interoperability anytime soon. We note, too, that most of these protocols represent very large and complex software implementations, where interoperability between vendors is measured in degrees.

Interoperability issues

We asked if two vendors' systems running the same standard protocol would interoperate. The chances are only about one in three that two systems would interoperate today "to a fair extent," functionally speaking, out of the box. That's not very reassuring to users.

The chances are improved to better than one in two, however, that some technical jury-rigging would allow them to interwork to a fair extent. The vendors optimistically project that in two years the chances that two systems will interoperate to a fair extent "out of the box" rises to two out of three.

Are today's leading voice-over-IP protocols the ones that will be the basis for interoperability in two years, or will still more new voice-over-IP protocols be needed? There are enough already - 80% of respondents say interoperability in two years will be based on today's protocols.

Is a two-year timetable toward interoperability reasonable? The vendors were asked: In what year do you expect most voice-over-IP equipment (greater than 50%) will be generally interoperable? The clear majority of them, nearly 70%, were split over whether this would occur in 2002 or 2003. That should help users who are tying their voice-over-IP purchase plans to the advent of interoperability.

We also asked vendors to name the main factors frustrating interoperability. The most cited factor, by 54% of respondents, is too many standards. The next most prevalent response, by 53%, is that marketplace pressure for product differentiation - that is, unique features and capabilities - is making interoperability difficult. Two other leading causes, both cited by 49% of respondents, are: "Certain leading vendors are not serious about interoperability," and "the voice-over-IP standards are too complex."

Latency is key

We also asked the respondents, most of them product managers, several questions about voice-over-IP performance and voice-over-IP voice quality. The most significant factor affecting voice quality is latency. As a follow-up question, we asked the respondents what they thought the maximum one-way, end-to-end latency acceptable to most people was (see chart).

In years of testing, we have found most people will begin to notice latency, even begin to be annoyed by it, as it rises above 150 msec. While 28% of respondents agree that 150 msec is the high end of most people's latency tolerance, one-third of respondents say that delays up to 250 msec are tolerable by most people.

At that point, which is akin to the latency of a voice call sent via a high-orbit geosynchronous satellite, callers would certainly be aware of the delay and probably also be annoyed by it. Clearly some vendors think consumers are willing to routinely accept anomalous voice-over-IP voice quality, presumably as a trade-off to other factors, such as low costs for phone calls.

Pros and cons

We asked the vendors to name the main factors that promote and propel voice over IP, as well as the main factors that inhibit or impede voice over IP's acceptance. In each case we offered a half-dozen check-box responses, plus space to write in "other" reasons (see chart).

Clearly the ability to move voice and data onto the same transport network is the most prevalent factor favoring and promoting voice over IP. Other sales points are derived from the consolidation of voice- and data-network management, and cost savings by reducing transport costs. Another major factor promoting voice over IP is that it enables new services, features and capabilities. Examples include unified voice and e-mail messaging, and various computer-telephony integration applications.

The main factors inhibiting voice-over-IP dominance are more varied. Lack of interoperability and the dearth of voice-over-IP expertise are two important factors - affecting availability of trained staff and even customer education. Too many standards and technological complexity were other leading reasons. The most prevalent write-in response was that lack of ubiquitous quality of service (QoS) continues to delay voice-over-IP deployment. Others cite the relative immaturity of voice-over-IP-based advanced features and functions, as well as large-scale performance and scalability.

Conclusion

Based on the survey responses, there is general agreement on several key trends regarding voice over IP. For one, most gear has been deployed by carriers and service providers, but shipments to corporations over the next two years are expected to balance that out. Also, as much as 8% of U.S. phone calls today are being carried to some extent in voice-over-IP form, and that will rise to a projected 36% in five years.

Interoperability, stymied by too many competing, complex protocols, should improve considerably within the next two years, the respondents agree. They note, too, that widespread QoS deployment needs to occur, and nifty new voice-over-IP features, functions and applications need to proliferate.