Beyond the backbone

There are two types of asynchronous transfer mode switch users: those who like them and those who don't. Carriers, which often hesitate to put ATM switches outside the core of their networks, usually fall into the first group, while Internet service providers tend to dominate the second.

Of course, there are exceptions. ATMNet, an ISP and reseller in San Diego, delivers service over an all-ATM switched network. But in general, the market is divided between the older, established telecom world and the newer, modern data world.

There is actually a third group of ATM switch users: the incidental users who don't care. Business customers that want high speed and service level guarantees-maybe in the form of transparent local area networks-aren't concerned with the type of switches used. Like the typical telephone customer who only wants a dial tone when picking up the phone, the typical ATM carrier customer only wants reliable service when transporting data between computers. Ironically, these indifferent users are actually driving the popularity and deployment of ATM switches.

What's it good for? Carriers and ISPs now buy ATM switches for one or two reasons: high speed and quality-of-service guarantees. But another reason beckons on the horizon, and some carriers are taking the first steps to get there: Telecom carriers are beginning to entertain the possibility of one network for both voice and data. That means migrating to a network using either ATM switches with voice capabilities or voice switches with ATM capabilities.

Vendors have noted carriers' affinity for consolidated networks, and some are beginning to offer appropriate switches. Siemens Corp. and Newbridge Networks' first joint project is an ATM core switch with voice SS7 capabilities that will be ready by the end of the year. Northern Telecom says it will add ATM capabilities to some of the switches in its DMS line. Other vendors are heading in a similar direction.

While analysts generally agree that any kind of consolidated switch is still a thing of the future, carriers and vendors are laying the groundwork. For instance, MCI announced several months ago that it would start bridging its voice and data switches with a service called Vault.

Initially, users will be able to use Vault technology for only a handful of applications, such as being able to talk to a service representative while on MCI's Web page. However, the connection between the voice and data world is more significant, representing a step toward a single, unified network. Sprint announced a similar service right around the same time.

The next step will be to integrate the management of the voice and data switches, according to Traver Kennedy, director of wide area networks for The Aberdeen Group, Boston. In the meantime, standard ATM switches are finally getting widespread acceptance and deployment. ATM switching has been a long time coming and was derailed several times. Remember 1992, the supposed year of ATM? Could it be that 1997 is finally the year of ATM?

Tim Burke, analyst at The Yankee Group, Boston, cringes when he hears questions like that. "The year of ATM is more like the decade of ATM. It's a revolutionary technology, but the deployment has to be evolutionary," he says. "The carrier commitment to ATM and purchases of switches happened to come at a critical time for ATM.

ISPs: The reluctant users The largest ISP is UUNet, first purchased by MFS less than a year ago and then again through WorldCom's buyout of MFS this past fall. The giant ISP uses ATM in three ways, and it employs ATM switches in one of those areas begrudgingly.

First, UUNet uses ATM inside its hubs to connect routers and switches in an ATM LAN, says Alan Taffel, vice president of marketing and business development. UUNet also uses ATM to interconnect hubs in five metropolitan areas in the country where one hub is not enough.

The third area is the one of most resistance. UUNet recently announced that it will implement Fore Systems' ATM switches across 20 long-haul links to create a nationwide OC-12 backbone (Figure 1). But Taffel does not mince words in saying UUNet is only using ATM because it needs the speed and there is no alternative.

"Our traffic load demands that we move to OC-12 speeds. In terms of connecting [Internet protocol] over OC-12, there is only one game in town at the moment, and that's ATM," he says. "This does not imply that ATM is our long-term strategy. It's a tactical move, not strategic. In general, we feel that ATM has fairly significant overhead, and we would prefer a frame-based solution." About one-third of UUNet's bandwidth gets chewed up by ATM overhead, which is not efficient or effective, he says. To get to an alternative solution, UUNet is not only turning to vendors but is also working with them. "As the largest ISP, we're really driving this," he says.

Some of the possible solutions are IP switching from Ipsilon, tag switching from Cisco Systems, and newer ideas like IP directly over Sonet.

"Everyone is still looking for the application for ATM," Taffel says. "We're using ATM for speed only. If you want to be at OC-12 and run IP, it's ATM or be on the bleeding edge with another solution.

Although UUNet is aggressively testing bleeding edge solutions, it has a commitment to its customers to carry traffic reliably. Still, talking to Taffel, one gets the feeling that as soon as another solid solution is standardized, UUNet will be at the front of the line (Table 1).

"Data folks are that way," Kennedy says. "They're not used to living within the levels of service and guarantees. The most common perspective is that when Cisco comes out with their migration path, the data people will move to ATM.

For Cisco, that means a double-edged sword. The company has many faithful customers who are ready to adopt its strategy, and the StrataCom acquisition has provided ATM switching expertise. However, the company risks losing key customers if its new product strategy is not satisfactory.

"Cisco clearly understands the importance," Kennedy says. "I just don't think they're ready yet.

Speed may not be the only reason ISPs are buying ATM switches, says Tom Nolle, principal of CIMI Corp. ISPs can also use them as "big, fat interface converters" to aggregate router traffic. One big T-3 interface to the carrier is cheaper than many smaller ones. "The press has misinterpreted ISPs using ATM as some grand preparation for the future. They, of course, aren't going to say otherwise," he says.

Switch cost is another factor for ISPs. Only the largest can afford ATM switches, says Liza Henderson, broadband consultant for TeleChoice, Verona, N.J. Small to medium-sized ISPs let the carriers buy the switches and then simply pay for ATM service, she says.

What it comes down to is a difference in the way individuals look at networks. Data-oriented companies such as ISPs build networks from the modems and routers on up. Telecom entities build from the core switches on down.

"I think the telcos are comfortable with ATM because it's a connection-oriented solution," says Taffel. "They are less comfortable with technologies like IP over Sonet, for instance.

Carriers: ATM underdogs? The interexchange carriers and the Bell regional holding companies, as well as their new competition, are indeed comfortable with ATM switches. When the cry for ATM began in the late 1980s, telcos took up the mantle. Since then, however, installation has proceeded at a plodding pace. Telcos have been the tortoise to the datacom industry's hare.

MCI, Sprint and AT&T, however, are the fastest of those turtles. They and other long-distance carriers have committed and installed ATM switches in the backbone to handle Internet traffic and to consolidate different networks in the future.

The competition the IXCs already faced, along with new competition allowed by the Telecommunications Reform Act of 1996, has forced them to look for cost-cutting measures, Burke says. Network consolidation is one way to do that.

Pacific Telesis, like the other RHCs, is looking "very seriously" at merging networks, although the demand for voice and data on a single network is still developing, says Jim Diestel, Pacific Bell director of advanced services. Right now, Internet voice is limited mostly to international use. However on the data side, Pacific Bell has already hooked up its first California general-purpose ATM network, created with Newbridge switches and a StrataCom-switched Internet backbone.

The differing views of ATM switching are the result of the different types of traffic that ISPs and telcos carry, Diestel says.

"We find that ATM is highly successful for us in the backbone. However, we recognize that we want to use ATM for much more than the Internet," he says. "Cisco's tag switching and Ipsilon's IP switching make more sense if you're only carrying Internet data traffic. How that will play out if voice and video become important on the Internet will have to be seen.

Pacific Bell uses StrataCom ATM switches in its network and could be well-positioned when Cisco releases plans involving routers and ATM switching.

Pacific Bell tested Bay Networks and 3Com as well as Cisco for ATM switch compatibility before the buyout, Diestel says. Cisco fared the best with the StrataCom BPX.

By using ATM switches primarily in the backbone, telcos are masking the technology when offering services to customers, says Henderson.

Robert Deaven, Bell Atlantic product manager for ATM services, agrees. Most of his customers don't care if they're using ATM switches, as long as they get reliable service.

For those customers, Bell Atlantic offers transparent LANs and other carrier-managed wide area services. Still, a few customers want to maintain control by buying their own ATM equipment.

"The number one drive for ATM is mixed media communications," says Deaven. "Today's frame relay, [switched multimegabit data services] and private-line services drive people toward ATM for higher speed and when they want to start mixing media." The rollout of low-speed ATM T-1 switched services will serve as a conduit to open up the ATM switch market outside the carrier backbone, he says.

Progress with switch interface standards has also helped. LAN switch users now have a defined frame relay interface, and the ATM Forum is expected to complete LAN emulation and multiprotocol over ATM standards this year. Those standards will make it possible to carry QOS overhead information end-to-end through the network.

They just want service The real fighting about ATM begins when the discussion moves outside the backbone to the end user. Many were discouraged after the extreme optimism of ATM's early days.

"There was all this hype about ATM and seamless LAN/WAN integration," says Burke. "People thought you could go home Friday night and on Monday have ATM LAN/WAN service.

Manufacturers and users of routers may be especially skeptical. Because a lot of money is tied up in those existing devices, end users will carefully weigh the benefits before making a new investment in ATM switching.

"But once you put ATM switches in one part of your network, it's only a matter of time before you move them further out toward the desktop," says Burke. "Someday we're going to find out a lot of people are buying ATM LAN switches."