Like many technologies, voice over IP so far has failed to live up to expectations. Part of the problem is that carriers and enterprises have billions invested in circuit-switched equipment, which has slowed deployments. And QOS is still an issue. But in times of emergency, VoIP could become an important release valve for an overwhelmed public network.

On Sept. 11, many Americans shared an experience that arguably was as gut-wrenching as watching the terror in New York City and Washington unfold on their televisions: Trying to get through to loved ones on wireline and wireless phones. Making a connection that day was a hit-and-miss proposition as call volumes overwhelmed the public switched network and wireless networks nationwide.

Conventional wisdom says both the wireline and wireless networks performed better than anyone had a right to expect given the circumstances. It also says that building out network infrastructure to the degree needed to ensure every caller got through on the first try would be completely unreasonable given the cost and time needed to do so.

However, defense and intelligence experts believe the U.S. will suffer another major terrorist attack sooner or later. Should that be true, thousands more callers will face the same frustrations as those experienced on Sept. 11.

There may be an answer to that dilemma: voice over IP (VoIP).

VoIP possesses a number of attributes that make it a potentially effective release valve for the public network and wireless networks in the event of a national emergency. For starters, the Internet was designed and constructed from the beginning to withstand nuclear attack.

“The Internet was created initially for [the Defense Advanced Research Projects Agency] to connect the Department of Defense to key universities,” said Brett Azuma, a former telecom analyst for Gartner Dataquest and now senior vice president of marketing for IP Unity, which provides carrier-grade services platforms. “It was designed so that if a node went down, you could route around it.”

While the public network and the Internet are redundant networks, there is a fundamental — and crucial — difference in their architecture, according to Mike Gaines, marketing director for VoIP signaling and service solution provider SS8 Networks. Because the public network is oversubscribed by design and has a great many lines plugging into a relatively small number of switches, it is particularly vulnerable in circumstances such as those that occurred on Sept. 11.

“That's the Achilles' heel,” Gaines said. “If a switch is taken out or overwhelmed, you're screwed.”

On the other hand, the Internet is a highly distributed network that has less dependence on any one piece of equipment, said Peter Briscoe, president and CEO of Convedia, a maker of VoIP enhanced services media processors. “There are no large switches and lots of routers and algorithms, so ‘follow me’ is much better compared to the [public network],” he said.

Brian Forbes, co-founder and vice president of business development for IP service delivery system provider Netrake, agreed: “The IP network has more intelligence to route calls. It's easier to work around node failure or congestion.”

The IP architecture is also much more versatile at the customer premises compared with a PBX architecture, according to Alec Henderson, manager of product marketing for Cisco Systems' voice technology center.

“You can unplug any IP phone from the wall, walk down the hall and plug it back into any socket, and the network will recognize that phone and route calls to it — as long as the network is still running,” he said.

Converging voice and data on a single packetized network would give enterprise customers and carriers greater latitude in determining how to manage the traffic that moves over their networks, said Dan Mangelsdorf, vice president of carrier VoIP marketing for Nortel Networks. “In a crisis environment, you could prioritize voice and de-prioritize Web activity and e-mail,” he said.

In addition, the compression algorithm that manages IP calls could be changed to allow more calls to go through, and bandwidth could be scaled based on demand, something that cannot be done with the circuit-switched network.

“The calls may sound a little muffled compared to how they normally sound, but not so much that anyone would notice,” said Jeff Dean, group product manager of convergent services for Genuity.

All of which is well and good, but only to a point. The single biggest factor limiting the potential of the Internet as a backup to the public network and wireless networks is a lack of VoIP termination points. Although VoIP is beginning to gain traction in the enterprise space, because of the cost savings it provides for long-distance calls it has gained little cache with residential customers.

Service providers also have been slow to embrace the platform. Consequently, VoIP-generated calls terminate at the public network most of the time, which makes them just as vulnerable as conventional calls when the public network has been overwhelmed or compromised.

Most observers acknowledge that VoIP has failed to live up to the hype. Industry research firm IDC, for instance, reported that international IP telephony retail minutes totaled 5.6 billion in 2001, compared with 110.7 billion circuit-switched minutes.

Less certain is why VoIP isn't further along in its development and adoption.

One theory is that customers — particularly business customers — still have reservations over the quality of a VoIP call. While the savings achieved by moving to VoIP are estimated to be as high as 20%, the lack of reliable QOS negates those dollar savings for many businesses.

“It's better than a cell phone, but when you pick up a [desktop] handset, you expect the quality to be like what you would receive from the [public network],” said Jim Mackey, director of product management for Intel's enterprise communications.

But Henry Sinnreich, distinguished member of engineering for WorldCom, scoffed at such a notion. IP networks have virtually no distortion or noise because they move traffic over a four-wire architecture, he said. Comparatively, the public network uses either two or four wires and thus is more vulnerable to frequency distortion.

“Remember when you were told not to order products online because you would be stolen blind?” Sinnreich said. “Voice quality is another lie told by those who wish to criticize the Internet, either because they're technically clueless or self-serving.”

Another hurdle VoIP must overcome is that carriers and enterprise customers are reluctant to invest in the technology because they already have poured a great deal of money into circuit-switched equipment, investments that need to be fully depreciated before they can consider widespread investment in a new platform. And the collapse of the competitive carrier market removed much of the incentive incumbents had to upgrade their networks.

“It's one thing to replace a thousand phones and another to roll out 65,000,” said Steve Joyce, director of network technology for NetIQ, which offers VoIP migration and management solutions.

Roland Soohoo, vice president of marketing for empowerTel Networks, believes he has a possible solution to this challenge: A convergence processor that through load-delay technology can process up to 1024 channels with just a millisecond delay between the input and output, eliminating the need for voice compression. In addition, through an embedded time-slot interchange device, the chip enables the simultaneous performance of TDM and IP operations.

“The result is a bridge between circuit-switched and IP networks without the wholesale switching of equipment,” Soohoo said.

Joyce suggested that vendors wishing to sell carrier and enterprise customers on the virtues of voice over packet concentrate on return on investment rather than features. “It's not enough to be cool. You have to show the IT managers that VoIP will save them a lot of money.”

However, Timothy Jasionowski, director of product incubation and strategy for Qwest Communications, said the inherent cost savings of IP is driving a “very aggressive deployment” of the platform into the core of the carrier's network. “This is a tough economy right now, and this is going to save the company a great deal of money, which is a good deal for Qwest's stakeholders.”

However, cool features that provide “intuitive management” of voice and data applications — such as visual voice mail — are the gateway to the future success of VoIP, predicted Jeff Gustafson, senior director of marketing for Vpacket Communications, which provides IP-based routers and management systems. Converged voice and data will move forward as more customers use enhanced services and eventually achieve a critical mass, he said (see figure).

VoIP FOR THE MASSES Forecast North America voice-over-packet revenues, 1999-2005 (in millions)
1999	258
2000	331
2001	404
2002	490
2003	606
2004	1072
2005	3577
Source: Probe Research

“VoIP will eventually achieve critical mass. It will be a momentum kind of thing,” similar to what happened when people started using e-mail and moved away from FedEx, Gustafson said.

The momentum will accelerate as more computer users upgrade to the Windows XP operating system, which has an embedded session initiation protocol client, said Michael Burrell, senior product manager of enterprise convergence solutions for IP and data services provider Equant.

Users who want VoIP will simply point and click to sign up, as they do for Internet access through AOL or MSN. Adding an inexpensive headset or microphone will complete the transition. When that happens, the lack of endpoints hurdle currently preventing VoIP from becoming a viable backup could well be cleared once and for all, Burrell said.

“That's when you will really see things take off.”