BOSTON (06/15/2000) - THE FUTURE OF PHONE? Voice-over-DSL offers cost savings--and considerable risk By Daniel Sweeney It's an axiom of the IT world: Don't muck with your phone system. Any given enterprise, can easily demonstrate horrendous losses when its phones go down for even a day. Any longer and your clients will likely assume you're out of business. So why would you even consider entrusting your local phone service to companies that in some instances didn't exist a year ago, touting technology that is largely without a track record?

According to the purveyors of that technology, namely voice-over-DSL (VoDSL), which packs multiple voice channels into a single copper line, price is the key. The industry claim for VoDSL--nearly impossible to verify, given the paucity of customers at present--is that local service costs will typically run 30 percent to 50 percent less than plain old telephone service (POTS). Most of the VoDSL vendors also offer long-distance telephone service and high-speed data access as further lures to new customers. Many can even provide extended wide area network connections over frame relay or asynchronous transfer mode.

As a result, the vendors are apt to further discount local phone service as bait for the total communications package.

It's also likely that VoDSL pricing will be consistently lower than POTS simply because VoDSL requires considerably less physical infrastructure than conventional local phone service. Existing VoDSL platforms can transmit eight to 16 voice channels simultaneously over a single phone line, and technologies exist that would easily permit a quadrupling of that figure. Within an existing in-building phone system, this means that adding a phone line or multiple lines simply involves adding a DSL modem.

DELAYS, DELAYS, DELAYS So far, however, voice-over-DSL has been among the service offerings of not more than a handful of the scores of DSL providers scattered across the country. Indeed, actual commercial deployments probably number no more than a dozen, according to Stephen Gleave, vice president of marketing for Los Gatos, California-based Jetstream, a manufacturer of customer-premises equipment for implementing VoDSL.

"The initial DSL service providers have concentrated on building high-speed data services," Gleave explains. "And they've chosen equipment accordingly."

And moving into voice requires features that most first-generation hardware doesn't provide. But, Gleave says, the revenue potential of local phone service (approximately $52 billion versus $8 billion for data services) provides a powerful incentive for competitive local exchange companies (CLECs)--such as Covad Communications and Rhythms NetConnections, which lease phone lines from local telephone companies and then resell the bandwidth to their own customers--to move into voice.

VoDSL also makes that move a real possibility for the first time. Despite the clear intent of the Telecommunications Act of 1996, competition for local voice traffic has scarcely appeared, primarily because of the dependence of most CLECs on leased capacity from incumbent service providers, such as BellSouth and US West. But VoDSL changes the whole equation. CLECs must still rent copper lines at an average of $20 per month. But now that $20 buys as many voice channels as the VoDSL technology will support--16 today but many times that in the foreseeable future.

The capability of VoDSL to leverage existing copper more efficiently may even force the regional Bell operating companies to adopt the technology. Robert Kelley, general manager for voice-over-DSL at Englewood, Colorado-based Rhythms, a nationwide provider of DSL data services, goes further, predicting a general replacement of POTS by VoDSL, at least in the business sector. "I see the [incumbent local exchanges] as being very vulnerable if they don't adopt the technology as well."

DSL GETS A VOICE The notion of using DSL for voice is hardly a stretch, since the technology was originally developed to support streaming audio and video so that the telephone companies could compete in the cable-television market without changing the physical layer of their networks. Subsequently, providers have used DSL to simulate T1 connections and to offer high-speed Internet access and virtual private network connectivity, primarily to small businesses and consumers.

DSL uses digital signal processing techniques to extend the information-carrying capacity of ordinary copper wire. DSL service providers usually assign only a single analog voice channel per line and use the additional bandwidth for data. But all or part of the data channel can also carry voice. Eight standard 64-kilobit-per-second voice channels require just over one-half a megabit of throughput--well within the capabilities of an average DSL connection. By using a single line to carry multiple voice channels, the carrier achieves significant cost savings for itself and, by extension, for its customers.

Transmitting voice over a DSL connection is a fairly trivial exercise.

According to Stefan Knight, director of product marketing for CopperCom, a Santa Clara, California-based supplier of VoDSL voice gateways and IADs, "There's no major technical hurdle to overcome--it's not like you're building a terabit router or something." Still, vendors must design specialized equipment for the purpose, including DSL access multipliers, which can treat voice separately from data; DSL modems that dock with small PBXs and generate the signals required by the telephone switch; and voice gateways that direct the VoDSL traffic into the public telephone system.

CUSTOMERS? SERVICE? The very recent arrival of such advanced equipment helps explain why some DSL providers have delayed their move into voice. But while VoDSL is primarily the province of relatively tiny companies (such as Rio Communications in Eugene, Oregon, Network Plus in Quincy, Massachusetts, and Picus in Virginia Beach, Virginia), at the moment a few nationwide DSL service providers, such as San Francisco-based Northpoint Communications and Rhythms, as well as Santa Clara-based Covad, are giving the technology a trial run. Such companies are exceptions, however. "The big guys are holding back until the technology is thoroughly proven," says Brett Sheppard, a former analyst with TeleChoice in Denver. "In my opinion, they're wise to do so, because it still has some problems, particularly with echo cancellation. We don't expect fully carrier-class equipment until the third or fourth quarter of this year."

The Antenna Group, a New York-based early adopter of Network Plus's VoDSL service, confirms Sheppard's misgivings--at least in part. "We've had ongoing problems interfacing the Jetstream subscriber unit with our PBX," reports former Technical Director Bill Meyer. Meyer has no plans to drop the service, however. "We're an Internet content delivery service, so we're willing to try new technology. Network Plus has been willing to hold our hand, and the service is much cheaper than what Nynex was offering."

PaintingsDirect.com, a New York City-based online art gallery, has been more fortunate with its Network Plus service. "We really like it," says John McLane, director of operations. "Voice quality equals that of ordinary phone service, and the cost is much less. There have been a few glitches, but you expect that in a new service."

Such mixed verdicts will undoubtedly cause many enterprise users to hesitate.

And surprisingly, one equipment manufacturer's representative--CopperCom's Knight--believes the wait-and-see approach is warranted. "People see the promise of tremendous savings, and they're skeptical--as well they should be.

We build to five nines reliability, but not everyone does." But Knight adds, "The incumbents are already buying equipment to stave off the CLECs. It doesn't matter who's offering the service; it's simply 16 times cheaper in terms of line charges. And that's why it will ultimately become the norm."

Dan Sweeney is a technology writer based in Burbank, California. He can be reached by e-mail at dswee34359@aol.com.

REVISIT IVR PHONES STILL DON'T LISTEN

Interactive voice response and speech recognition make slow progress By Fred Hapgood If life had any logic to it, computers would have been developed by the telephone industry. Instead they grew up everywhere but. First they were in the military, then in universities and finally in internal business applications such as payroll and accounting. In fact, telephony and information processing grew up isolated from each other, with different technical perspectives and vocabularies. Few engineers understood both.

As a result, today's great communication technologies have found it diffi-cult to...communicate. Computers could phone us after a fashion, using prerecorded words or voice synthesis, but managing a computer from a phone has always been problematic.

In February 1991, CIO ran an article declaring that this wall was about to come down, and the age of "anytime, anyplace" computing would begin. We based our optimism largely on a rash of agreements by industry standards groups. These agreements, for instance, associated certain functions with specific touchpad actions: Press 1 to listen, 2 to save, 3 to erase and so on. Vendors hoped that this standardization would make it easy for users to remember the right key to press at the right time and therefore extend the voice-mail model to everything we think of today as e-commerce: perusing catalogs, placing orders, requesting customer service and more.

But interactive voice response (IVR) never spread far from its primary niche-- voice mail in corporate calling centers. IVR didn't fail for the usual reasons technologies stall: misconceived applications, inadequate investment, poor performance or overestimated demand. In this case, the target applications were well understood, the machines worked to specs, and the potential market--anyone who did business over the phone--was already huge and getting larger every day.

The problem was us--humans weren't up to remembering sequences of key presses, standards notwithstanding. (It didn't help that phone makers also had an annoying practice of building their touchpads into the handset, forcing users to keep pulling phones away from their ears to enter commands.) In theory there was an alternative: speech recognition. Unfortunately that technology had the exact opposite problem of IVR. Humans loved it, but machines couldn't handle the job. Even after decades of research, speech recognition still couldn't deliver acceptable performance in 1991. "There are thousands of ways to make a phoneme [speech's smallest phonetic unit]," explains Bill Meisel, president of TMA Associates, a Tarzana, California, consulting company specializing in speech recognition. "The sound that we hear is a product of the geometry of the voice box--which is different for every person and affects the phonemes just spoken, the ones that will be spoken next, how much noise there is in the environment, the speaker's meaning and his or her age, physical and emotional state, and many other factors." Up to the early '90s, the difficulty of building machines capable of handling this swamp of verbiage had pretty much confined speech-recognition technology to Star Trek.

Meanwhile, a huge market was left waiting. But during the last decade, speech-recognition companies made significant investments. Field-workers compiled tens of thousands of speech samples. Theoreticians worked to improve the algorithms that differentiate speech from noise. And programmers developed ways to make use of the faster processors that kept arriving. Circa 1997, functional speech products, such as Dragon System's NaturallySpeaking, IBM Corp.'s ViaVoice and Wildfire Communications' WildFire, began to appear.

These products let individual users--often after a lengthy training process--dictate text and commands to their desktop PCs. But more recent products can work with multiple speakers and can be used over networks, including the Internet. For instance, Universal Studios, the entertainment giant in Universal City, California, has replaced its internal voice mail directory with a speech-recognition system (by Philips Speech Processing) that recognizes more than 10,000 names. TelSurf Networks is selling an "audio browser" that uses recognition to navigate through a "speech portal" that delivers audio e-mail, news, stock quotes, radio programs, calendar entries and more. Several companies, including Brooktrout Software and InterVoice-Brite, now sell tools that allow Web navigation through speech--replacing the keyboard and mouse. Users either speak into a microphone and deliver commands using voice over Internet protocol, or they call a number on a second phone line and speak over that.

Nearly 10 years after our story, the merger may finally be at hand--but it won't be easy. Carl Strathmeyer, director of strategic marketing at computer telephony vendor Dialogic, points out that a computer has to do more than recognize phonemes; it has to understand the semantics as well. It has to know what the decoded words mean. To aid that task, people must vocalize clear, simple, unambiguous and invariant requests and responses. Strathmeyer observes that humans may resist such discipline, suggesting that the technology might run into the same social hurdles as touchpads--even if the technology is ready to deliver.

Now you have to manage the phones. What's next, the watercoolers? Send your complaints and comments to Christopher Lindquist at et@cio.com.

PREDICTIONS BUSINESS PROCESS INTEGRATION GET FAST OR GET LOST Get ready for real-time. By 2004, more than 70 percent of the e-business applications your customers see will need to offer near-instantaneous integration with your back-end systems, or they won't be doing their jobs, according to Stamford, Connecticut-based Gartner Group.

The reason is simple--slow response will equal bad customer service, says Roy Schulte, vice president of Gartner's application integration and middleware practice. As customer expectations increase, their willingness to wait for feedback after placing an order, asking a question or making any other request will diminish. Users will also become frustrated if they find they can get faster service over the phone than they can with a browser--putting added pressure on your call center that the Web was supposed to help alleviate.

As a result, the days of being able to accept an order online, only to have it faxed to a distributor or rekeyed into a back-end order tracking system will quickly pass, Schulte says. And, he adds, companies that want to update their systems to provide faster response need to look beyond hard-dollar-cost justifications. Although it's true that moving to an automated system may let companies eliminate some clerical help, the real money will come as a result of the impact the faster system will have on the business.

For instance, Schulte notes, if a company can reduce its order-processing error rate from 7 percent to 4 percent by updating systems, that alone could save millions of dollars a year. Speedier response can also mean more repeat business as well as the chance to recognize revenue more quickly: The faster your orders go out the door, the sooner you can charge customer credit cards.

And that means more money on your bottom line. -Christopher Lindquist UNDER DEVELOPMENT BETTER SOUND TELECONFERENCING GOES STEREO Over the years, the telecommunications industry has explored many technological frontiers, but the introduction of stereo sound hasn't been one of them--until now.

Bell Labs researchers recently made a breakthrough in audio technology that makes it possible for the first time to deliver full-duplex stereo sound over an audio or video teleconference. The technology works in an Internet protocol (IP) environment and can integrate into almost any IP-based teleconferencing system.

Stereo teleconferencing gives participants a better way to hear and comprehend multiple streams of conversation, says Gary Elko, research supervisor in the Murray Hill, New Jersey-based speech research department of Lucent Technologies Inc., Bell Labs' parent company. "It supplies the 'cocktail party effect'--the ability to listen in on a particular conversation when there are a bunch of conversations going on around you." He adds that stereo technology is well suited for use in an IP environment. "With IP's audio delay people tend to talk all at once. Stereo will help users sort out the different voices."

To make the technology work, Bell Labs researchers had to tackle a complex problem called "stereo acoustic echo cancellation." Without echo cancellation, conference participants would suffer the screeching feedback that occurs when a microphone picks up sound from a nearby audio speaker, forming an acoustic loop.

Current echo-killing schemes rely on a one-channel-cancellation technique, which identifies a single acoustic path between the speaker and the microphone.

This approach doesn't work in a stereo environment because the sound travels along multiple acoustic paths between the speaker and the microphone. To overcome this obstacle, the researchers developed new algorithms that modify the stereo signals in a way that allows for correct audio-path identification without harming the stereo sound.

Stereo conferencing technology has significant potential, says Francie Mendelsohn, president of Summit Research Associates, a Rockville, Md.-based kiosk and telecommunications research company. "It could be kind of cool," she says. "It's not something people have been asking for, but if it lives up to Lucent's promises, it could turn out to be very popular."

Bell Labs will add the stereo teleconferencing technology to a number of Lucent teleconferencing products, including Lucent Collaborative Video, an IP-based system for PC users. Lucent also plans to license the technology to other companies. -John Edwards NEW PRODUCTS CHAIN CHAIN CHAIN Business-to-business connectivity got you bugged? E-Sync Networks may be able to help with TotalChain, the company's XML-based supply chain tool. The product claims a variety of features, including the capability to operate via a browser, fax or e-mail interface; client/server authentication for security; online billing and transaction tracking; warehouse and transportation fulfillment; and integration with applications from Baan, J.D.

Edwards, PeopleSoft and SAP. E-Sync can also host the TotalChain system on its own servers, reducing many hardware and implementation costs. Pricing varies depending on the application. For more information, visit www.e-syncnet.com or call 203 601-3000.

WIRELESS TORTURE TEST The power of a cell phone increased dramatically with the introduction of the Wireless Application Protocol (WAP). No longer just voice--and sometimes e-mail--communicators, WAP-enabled phones become portable, wireless data access devices. Now Segue Software offers a way to make sure that WAP applications will work as expected without time-and-money-wasting trial and error. SilkPerformer lets you test the performance of multiple requests to applications on a WAP gateway or Web server prior to actual deployment. Pricing starts at $19,500. For more information, visit www.segue.com or call 781 402-1000.

AN EYE ON YOUR NETWORK As your network becomes more mission critical, keeping it running smoothly is increasingly important. To help prevent problems before they start--and identify causes when they do--MediaHouse Software recently unveiled version 6 of its ipMonitor network monitoring package. IpMonitor installs as a Windows NT service that you configure to inspect network services or devices on a regular schedule. If ipMonitor detects failures or service degradations, it can alert network administrators via a variety of means, including phone, pager, e-mail and network broadcast. The product can also automatically restart services, reboot systems and activate external recovery tools. IpMonitor sells for $695. Download a free trial, and get more information at www.mediahouse.com or call 819 776-0707.

MORE SPEED, LESS EFFORT Networking newcomer Alacritech (founded by Larry Boucher of Adaptec) claims it can offer gigabit-class performance from your servers without overhauling your network infrastructure. The company's Server Adapters combine two- or four-port 10/100Mbps adapter cards with an onboard processor that can significantly reduce the load on the server's CPU, speeding response times and increasing overall throughput. According to Alacritech, other adapters can absorb as much as 100 percent of a CPU's processing power to handle network requests. The Alacritech Server Adapter, however, can reduce that load to 23 percent or less, the company claims. The two-port adapter costs $299. For more information, visit www.alacritech.com or call 408 287-9997.

MAKE A WISH If your support organization is swamped with requests and can't seem to dig itself out, Wishbone Systems may be able to help. Wishbone WorkManager 4.0 lets support departments manage requests through a graphical interface that provides instant feedback on the current status of tasks and resources. Managers can quickly assign tasks to personnel either manually or by using automated scheduling. Support staff can also request tasks and have them assigned automatically according to their availability and expertise. Managers can view customized reports that provide feedback on the support operation.

Pricing is $10,000 for an 11-person work group. For more information, visit www.wishbonesystems.com or call 201 541-7000.

PROCUREMENT POSSIBILITIES The business-to-business e-commerce arena is rapidly reaching standing-room-only capacity, though there always seems to be space for one more contender. Vsource.net says it will take on the big players--such as Ariba and Commerce One--by offering a simpler, more cost-effective procurement solution that can scale anything from small businesses all the way to global enterprises. The company's system--called the Virtual Source Network--is browser-based and reportedly integrates with common back-end software, including that from Oracle Corp. and SAP AG. The service is free for suppliers and starts at $25,000 for customers to establish a pilot program (with guidance from Vsource). For more information, visit www.vsource.net or call 805 677-6720.