★ wanayoo — archive 1999 http://data.com/tutorials/voice_over_atm.htmlNouvelle recherche | Portail wanayoo
CMP's TechWeb Click here to visit Nokia


Data Communications
Search Data Communications

 EDITORIAL CALENDAR 2000

 Browse By...
 Technology/Topic
 Vendor
 Back Issues

 Content
 LAB TEST CENTER

 TECH TUTORIALS


 STANDARDS WATCH

 GLOBAL NETWORKS

 N.P.N.

 PRODUCT LEADERS

 INDUSTRY PIPE

 Visitor's Center

 FAQs
 Contact the Editors
 Free Newsletter
 Subscriptions

 Marketing Services
 Industry Front & Center

 Custom Publishing
 Vendor Strategies

 CommWeb Sites
 Computer Telephony
 Call Center
 Teleconnect
 ISPs
 Tele.com

TechWeb Sites
 Byte.com
 CMPmetrics
 Data Communications
 File Mine
 InformationWeek
 InternetWeek
 Network Computing
 Planet IT
 TechShopper
 TechWeb News
 Tele.com
 WebTools
 Winmag.com

Data Communications: Network Analysis

February 1996


By Kieran Taylor

Voice Over ATM: A Bad Connection

Data-driven specifications are making for inefficient voice service

[ Talking the Talk ]

Magic bullet or misfire? That's what net managers who turned to ATM to carry voice and data over the same network must be wondering. The technology of the future is proving less efficient (and more expensive) for voice traffic than time-division multiplexing (TDM) over leased lines, which has been around for a couple of decades.

The problem lies with the ATM spec itself. The ATM Forum currently mandates that voice be transmitted as CBR (constant bit rate) traffic--a method that forces customers to reserve bandwidth for voice even when they're not actually sending it. Sending voice as VBR (variable bit rate) traffic is the obvious alternative, since it permits net managers to allocate voice bandwidth on an as-needed basis--and possibly cut voice call bills in half, according to some consultants.

Tro uble is, VBR for voice isn't standard. Although the ATM Forum is currently looking at ways to write it into a revised ATM spec, bickering among forum members could delay any action on that front for at least a year. Some vendors are taking advantage of the situation, leaping into the breach with so-called prestandard VBR equipment. The downside to such gear? It's proprietary, and that leaves customers in a voice-over-ATM double bind: Invest in standards-based but inefficient CBR products, or sacrifice interoperability for savings by going with VBR offerings.

STANDARD SUMMARY

The ATM Forum has generally concentrated its efforts on developing standards for data transmission. "The Forum isn't that interested in voice processing," says Tom Nolle, president of consultancy CIMI Corp. (Voorhees, N.J.). But users are--to a degree that Nolle says is "surprisingly high."

The forum only started work on voice transport in 1993, and it wasn't until April 1995 that the VTOA (Voice and Telephony Se rvices Over ATM) working group published its first document, which contained the unstructured circuit emulation and structured circuit emulation specs. Unstructured circuit emulation maps an entire T1 (1.544-Mbit/s) circuit to a single ATM virtual circuit (VC), thus limiting it to point-to-point applications. Structured circuit emulation allows switches to map individual 64-kbit/s circuits in a T1 line to ATM VCs, and it can be used for point-to-multipoint connections. An enhancement that will allow switches to add and subtract 64-kbit/s circuits automatically in response to traffic levels is due this summer. Each requires voice to be treated as CBR traffic.

At first glance, that doesn't look so bad, since CBR guarantees that bandwidth will always be available for voice calls. The only problem is that it does so by establishing a VC with a fixed amount of bandwidth for each voice connection. Customers thus have to pay for the bandwidth whether voice is being sent or not (see the figure). VBR, on the o ther hand, lets the bandwidth be used for data when no voice traffic is being sent.

Constant vs. Variable: How Voice is Carried


The current ATM standard specifies that voice be carried as CBR traffic, a less-efficient method than time-division multiplexing. Because a virtual circuit is established with a fixed amount of bandwidth for each voice connection, users have to pay for the bandwidth whether voice is being sent or not. So-called prestandard gear now being sold allows voice to be carried as VBR traffic--the only problem is that it locks customers into a single-vendor solution.

Ironically, the problem affecting ATM CBR also plagues TDM nets, which use multiplexers to transmit voice and data over point-to-point leased lines. But TDM nets still manage to use that bandwidth more efficiently. The overhead incurred when encapsulating voice into ATM's 53-byte cells equates to a bandwidth hit of 74 kbit/s--10 kbit/s more than that incurred in a TDM network. The number of voice circuits a standard ATM switch can fit onto a T1 line is thus limited to 21, compared with 24 for TDM gear.

It's this inefficiency that would offset any cost benefits to using ATM--even if it were to be priced lower than T1 TDM service. Most consultants, however, think any difference in rates is minimal (carriers don't publish their tariffs, so it's tough to get an exact read).

There are now about a half dozen vendors shipping products that implement the ATM Forum's voice standards. Some of these, like General Datacomm Inc. (Middlebury, Conn.) and the Lightstream division of Cisco Systems Inc. (San Jose, Calif.), handle only unstructured T1 interfaces. Other vendors, such as ADC Kentrox (Portland, Ore.), Fore Systems Inc. (Warrendale, Pa.), and Newbridge Networks Inc. (Herndon, Va.), support b oth unstructured and structured T1-to-ATM interworking.

VARIATIONS ON VOICE

But users are growing wise to the drawbacks of sticking with the ATM standard, and they're making their dissatisfaction clear.

"Users on the ATM Forum's Enterprise Network Roundtable group say bandwidth savings are paramount," says Peter Chase, chairman of the VTOA group responsible for developing voice ATM standards. "Their view is that the cost differences between CBR ATM and conventional TDM are trivial."

To give net managers the fiscal incentive to move to ATM, the forum is now looking at ways to make VBR available for voice. Silence suppression and voice compression could also be a part of a new spec, and that would enable net managers to make even greater use of bandwidth. But the work is going slowly, mostly because members can't agree on the best way to proceed. "It's very unlikely that the current proposals will be ratified as a standard within the next year," Chase says.

Nolle sees a problem with the structure of the forum itself. "The political weaknesses of the ATM Forum are becoming apparent," says Nolle. "It may be in the interest of vendors that lack voice support to delay such a specification."

Nevertheless, there are a couple of approaches under consideration. One is based on AAL (ATM adaptation layer) 6, which is now a proposal in ITU Study Group 13. Although AAL 6 defines VBR service for low-bit rate voice traffic between a wireless base station and mobile switch center, Chase says it could be parlayed into a new voice standard.

One thing in AAL 6's favor is that it already defines voice compression and silence suppression. Yet settling on standard forms of these features could itself prove difficult. That's because ATM equipment vendors are implementing a variety of schemes, using algorithms that compress a 64-kbit/s channel into 32, 24, 16, or even 8 kbit/s of bandwidth. Silence suppression techniques--which ensure that none of the pauses that make up half of a conversation are treated as traffic--are no less numerous.

HEARING VOICES

Some vendors have made the end-run around the ATM Forum and gone directly to users. IBM, Northern Telecom Ltd. (Mississauga, Ontario), and Stratacom Inc. (San Jose, Calif.) are responding to the clamor for VBR gear by selling nonstandard switches. Newbridge and Premisys Communications Inc. (Fremont, Calif.) plan to follow suit.

These products use proprietary AALs that enable voice to be carried as VBR traffic, and they come with compression and suppression features. It's a combination that has corporate customers envisioning some big savings, since voice accounts for as much as 60 percent of their communications traffic.

"It costs companies between 12 and 19 cents per minute to send voice over a T1 network using TDMs," says Nolle. "If a company has enough data traffic to fill all the gaps between intermittent VBR voice transmissions, and they use compression, voice will cost 7 to 8 cents a minute." He adds that even those carriers connecting voice calls over VPNs (virtual private networks) charge 10 cents a minute--20 percent to 30 percent more than private networks using VBR voice.

The proprietary AALs used by IBM, Northern Telecom, and Stratacom are similar to the AAL 5 spec for dynamic data transfer, with one exception: They furnish the timing synchronization that's required for transmitting delay-sensitive voice calls. The switches also can convert traditional PBX signals into the Q.2931 signaling used on ATM networks, and this helps prevent ATM voice circuits from being triggered when they're not needed.

Kenneth Ledford, director of telecommunications for Ashland Inc. (Russell, Ky.), an oil company, says that VBR, combined with dynamic compression and suppression of voice traffic, was the reason he bought a Magellan Passport switch from Northern Telecom. "The voice quality is very good over ATM," he adds, "and I have a solution that's more flexible than TDM."

But because they' re not based on ATM, these VBR switches can't work together. And that's not the only risk in buying proprietary gear. Some ATM switch vendors warn that users may have trouble upgrading prestandard switches to work with the voice VBR spec when it's finally ratified.

"Standards for VBR-based voice are quite fluid, and there's a good chance that prestandard implementations will require hardware upgrades to meet forthcoming standards," says Bill Miller, director of product management at General Datacomm. Vendors of voice VBR switches deny this claim, insisting their switches will get by with software upgrades.

Of course, letting an ATM carrier worry about the upgrade is one way to go. Net managers who want VBR services without the potential upgrade hassles can opt for WAVE (Wide Area Voice Exchange) from MFS Datanet Inc. (San Jose, Calif.) (see "Consolidated Service: An Integrated ATM Network," October 1995). Corporations using dial-up services for voice and frame relay for data can save 14 percen t to 25 percent by switching to WAVE, according to the carrier. As of now, MFS Datanet is the only carrier offering this kind of service.


Illustration by Warren Lin

[ Home ]

[ Registration | Subscriptions ]
[ Contact Us | E-Mail ]



Home Contact Editors Lab Tests Registration Tech Tutorials
Buyer's Guide Global Networks Opinion / Columns FAQs Subscriptions

CMPnet Click Here for Computer Associates