Feb. 1, 2003 Issue of CIO Magazine | Emerging Technology

Net Talk Gets Real

Voice over IP isn't an if; it's a when

BY FRED HAPGOOD

On Sept. 11, 2001, the phone company's central switching station serving the headquarters of New York City's Department of Sanitation (DSNY) was crushed by the collapse of the Towers. The very agency with line responsibility for the Herculean cleanup effort that would soon be required suddenly found itself unable to communicate with its offices and personnel. Even worse, DSNY headquarters soon learned it was on its own so far as restoring service was concerned; its telecom provider—Verizon—had its own wounds to look after. But the Sanitation Department's commissioner made his needs clear to MIS Director Steven Stam: He wanted his telephones back immediately, whether it was possible or not.

Fortunately Stam had a few functional assets, including a reasonably robust LAN and the knowledge that there was a fiber data line running through the building that did not terminate at the destroyed Verizon facility. Strictly speaking, the fiber didn't belong to his department (it was leased by the Department of Health), but these were unusual times ("I begged, I borrowed, I was accused of stealing," Stam remembers), and by Monday, Sept. 24, he had started to lash those together to support a voice-over-IP (VoIP) network—using the LAN to carry phone calls and support a gateway into the public switched telephone network (PSTN). By the following Monday he received the go-ahead for the rollout, and the new phones began to ring one week later. With technical help from Dimension Data (a networking infrastructure services company in Reston, Va.), Stam had 285 VoIP phones running throughout the department's headquarters. Today, with 600 phones in three buildings on the new system, Stam is beginning to refocus on more traditional IS issues, such as tracking the reduction in costs. "Fifty percent of our phone calls are internal," he says. "The savings we get from putting those calls on the network are already substantial."


It is widely accepted that sooner or later VoIP will sweep the board.

It is widely accepted that sooner or later VoIP (also widely referred to as IPT, for Internet protocol telephony) will sweep the board, offering the promises of simpler management, lower costs and voice-application integration. The analysts predict it. The hardware and software vendors promote it. Even the telcos have begun to get in on the act. However, even the most enthusiastic advocate for the technology has to concede that at least for the moment, old-fashioned switched connection telephony is pretty good at its core mission: quality calls at high levels of reliability. VoIP features such as point-and-click dialing, integrated voice mail/e-mail management and CRM-to-caller-ID integration would be nice to have, no doubt—but none are yet mission-critical.

Further, VoIP is inherently demanding on the underlying network. Packet delays and losses that no one would even notice during Web surfing can make voice applications unusable. At the same time, telephony is perhaps the most important infrastructure service. From a network services point of view, supporting VoIP means becoming extremely dependable about delivering very high levels of performance. Moving to such a demanding technology can mean lots of network upgrades, acquiring smarter traffic monitoring tools and practices, organizing end user training sessions, and managing the delicate business of integrating two service groups (telecom and data services), each with its own culture. Timing such a labor-intensive step can be a difficult call.


One Toe at a Time
To date most VoIP transitions seem to be triggered by major changes in the corporate environment, such as a significant divestiture, acquisition or relocation. Some companies, however, are finding alternate paths: incremental transition strategies that move the enterprise toward VoIP one step at a time, minimizing up-front costs and risks while maximizing education.

FedEx Freight, headquartered in Memphis, Tenn., is the part of FedEx devoted to the delivery of fairly small ("less than a truckload") units of freight between businesses. By nature it is a highly distributed operation, with 335 facilities scattered around the United States. Jeff Amerine, managing director of Communications Network Services, says that when the impending arrival of VoIP became obvious, he decided to concentrate on transporting calls among PBXes over the company WAN, putting desktop telephony and other end user applications on hold. About two years ago he set up a series of lab tests and field trials. Once those were completed, he defined a group of offices that were smaller than average ("Where you could cut your teeth with less risk," he says) and whose PBXes were nearing the end of their useful lives. He equipped the more than 30 offices that met both criteria with a VoIP-enabled PBX system from a large global manufacturer and watched WAN traffic behavior for six months. Then he rolled out the service to another 40 locations. Amerine says he expects to have all FedEx Freight's facilities sending phone calls over the WAN by 2004.

He notes that the big surprise and lesson of his experience is the large number of tweaks—mostly hardware and software upgrades—needed to bring network quality of service up to the level required for voice. He adds that these tweaks are sufficiently sensitive to local features that an IT department would be well-advised to develop the necessary network expertise in-house. And incremental VoIP introductions give the IS department the time to master those skills.

Even companies with considerable experience with VoIP move cautiously into this new terrain. Halliburton Energy Services, the Houston energy giant, has been experimenting with WAN-based VoIP since 1996. (Currently about 15 percent of the company's more than 100 sites can place calls over the Halliburton WAN; there would be more, but VoIP is restricted by law in many of the countries where Halliburton operates.) In late 2000, Andy Knight, who is responsible for Halliburton's network communications (his actual title is Chief Gadget Guy), decided to begin moving to IP telephony. His first step was to install an experimental system that ran in parallel with a conventional PBX and carried calls only among members of a single office (with no PSTN gateway). He used this system to support experiments with features and build bandwidth utilization models.


From a network services point of view, supporting VoIP means becoming extremely dependable about delivering very high levels of performance.

By late 2001, Knight had gained enough experience to move the technology into a production setting. Once again, he picked the site that would deliver the most education for the least risk, specifically Landmark Graphics' (a wholly owned Halliburton subsidiary) London office, which was then moving into a new building. "We had all the business groups there but in a smaller package than in Houston," Knight says. (The London installation came to about 150 phones.) Like Amerine, he found that VoIP networks are picky about where they choose to work. "Laying a cable over a light might not affect performance in a conventional network, but even low levels of electrical interference can be a problem for voice," he says. Knight also agrees with Amerine that companies need to have their own expertise on the technology: "VoIP is still too new to expect contractors to know all these details themselves."

When Landmark moved to new offices this January, Knight provided desktop telephony services for procurement, IT, and shipping and receiving, but drew the line at moving senior management onto the system. "Given the value of the people and processes involved, we didn't feel we could do justice to an IP installation at the moment," he says, though a switchover to an all-VoIP system is still scheduled during the next 18 months. Meantime, Knight is moving on another issue just as incrementally: introducing a small number of "softphones" that ring right on laptops or notebooks. "You can check into a hotel, plug in the laptop, start up the phone, and so far as anyone can see, you're right in your office," he says.


Building on Voice
Some systems offer structural points of entry. Recently Bruce Elkington, CIO of Overlake Hospital in Bellevue, Wash., began thinking about the expense of installing conventional telephony in a building Overlake was planning to erect. VoIP seemed a logical alternative, so Elkington looked for a controllable installation that would give him some experience with the new technology. At one point he heard that the nurses on the general surgical unit were looking for cordless phones to supplement the unit's noisy and inefficient paging system. Elkington knew a wireless data network was already available (and used to serve the data devices carried by clinicians making their rounds), so he decided to take advantage of it to carry wireless VoIP.

His experience reinforces the points made by other VoIP implementers: 1. Network provisioning and configuration requires lots and lots of fine tuning; 2. Installation partners and contractors seldom know as much as they think they do; and 3. Once the system is working, it works very well indeed. "It just took a few days before everyone said, We want one of those things too," he says. Today Elkington is planning an expansion to 150 VoIP phones and is quietly confident that he is going to save a lot of money for his employer when the new building goes up without need for telephone cable. But he strongly endorses the strategy of defining an "educational" VoIP project before jumping into a major makeover. "If you don't find a way to get comfortable with the technology first, you're going to end up with a tin cup," is how he puts it.

These examples illustrate ways to design an incremental strategy: smaller versus larger offices, peripheral versus central enterprise functions, IS versus the rest of the company, WAN versus LAN (or vice versa), limiting the number of initial users, picking the offices with the oldest equipment or by capitalizing on a preexisting hardware base.

The particular strategy matters less than having one. The reason Steven Stam was able to work so fast that September, for instance, was that he already had run a VoIP project for months—in his case, a system that ran off the department LAN was distributed to headquarters but ran parallel and redundant to the PSTN system. People could use it as they pleased. And as they drifted on to the system, Stam was able to cut his teeth on VoIP.

Stam launched his experiment because of trends in the underlying economics. "I could see that prices kept coming down on the data side and not coming down on the telecom side," he says. But because he had done so, he was ready when his world changed in the most unexpected way. It's a useful lesson in this era, when the world might change as much for any of us at any time.


Fred Hapgood is a freelance writer based in Boston. He can be reached at hapgood@pobox.com.

 

Back to Whats New or Back to Home Page.