As 2002 begins, a handful of big data switch vendors are readying announcements of new monster multiservice switches they claim will help carriers revamp their core networks.

The new gear is based on – drumroll – existing Asynchronous Transfer Mode (ATM) technology.

Hold on: Isn't ATM supposed to be dead? (See ATM: Over and Out?.)

Far from it. Since the downturn in capital spending began, carriers have been looking for ways to extend their existing networks, and that means building on ATM and frame relay installations, at least for the foreseeable future (see Carrier Survey). The ATM switches, the demise of which was predicted in 2000, have gained a whole new lease on life.

But these new products aren't your father's ATM switches. Instead, they represent another big step up in capacity, and they incorporate extra functions aimed at helping carriers consolidate their current network infrastructures.

The upshot? Instead of withering away, ATM core switches are being augmented with capabilities that carriers think will enable them to preserve their existing network investments while ensuring a smooth migration to packet-based, optical core networks in future.

Suppliers of the new ATM gear include established players such as Alcatel SA (NYSE: ALA; Paris: CGEP:PA), Lucent Technologies Inc. (NYSE: LU), Marconi PLC (Nasdaq/London: MONI), and Nortel Networks Corp. (NYSE/Toronto: NT), as well as startups like Équipe Communications Corp.. In addition, Cisco Systems Inc. (Nasdaq: CSCO) offers its own multiservice switch, while hedging its bets on whether routing will predominate in the ATM core.

Where's it all headed? Can these new products fit the bill carriers have written? This brief report takes a look at the plans, hopes, and potential pitfalls of the next-generation ATM-based core switch market.

Page 2: It's Baaaaack!

Page 3: Where Is It?

Page 4: Where's it Going?

As this went to press, the leading multiservice switch vendors all confirmed plans to release new products shortly. Here's a chart of the highlights:

{Table 1}The information that's available about these new offerings raises more questions than it answers. For one thing, vendors are already starting a slap-fight over the merits of their unavailable gear relative to the unavailable gear of their nearest rivals.

Some vendors, for instance, are claiming that their products feature "real" switching speeds based on backplane capacity, as opposed to the internal switching matrices claimed by rivals (see Marconi Unveils Big Switch/Router). Others tout the amount of real estate in chassis as a differentiating factor. Équipe, for instance, criticizes Alcatel's multishelf array as bulky and costly to maintain (see Alcatel ATM Switch Steps Up).

"Everyone has their own spin to put on things," says Jim Lawrence, program director at Stratecast Partners. "Everyone plays games with numbers."

Clearly, the problems involved in migrating ATM to packet-based optical networks are complex enough that no single solution can be expected to emerge. And only time and testing will tell which ones rise to the top.

Even if there are several solutions that ultimately prove out, the timeframe for getting all the pieces in place will probably be extended over years, not months.

Still, the move is on. "There will be a gradual network migration from circuit to packet," says Chad Dunn, a cofounder and director of product management at WaveSmith Networks Inc.. "And it will start from the core and move out."

The evolution of ATM multiservice switches points to a key industry shift towardaddressing the needs of incumbent carriers, which typically run all of theirdata services (Internet and frame relay) over ATM infrastructure.

This is a fairly new development. Up until last year, a variety of carriers, including CLECs, were targeted by switch vendors as key customers, eager to install new equipment for their networks. It was thought that ATM and frame relay networks comprised a legacy network that incumbent carriers would soon replace with all-optical gear populated with Internet Protocol (IP) traffic.

Then came the bottoming out of the inflated "dotcom" market and the subsequent downturn in fortunes of newer carriers – not to mention the downturn in capital spending by incumbents.

Now the priorities have changed. "[Carriers] have been making money from frame, voice, and private-line networks, and they're not giving that up, even though they know new services are lurking on the horizon," says Dunn of WaveSmith.

Dunn and other vendors say carriers have decided to forego network overhauls that would require them to ditch their existing infrastructure in any way. Instead, they're prepared to maintain multiple ATM and frame relay networks alongside IP ones until they are compelled by demand to move wholesale to packet-based optical networks.

In response to this, vendors are creating the multiservice switches that are profiled in this report. These products are meant to work both ends of a carrier's present data networking setup – the connection-oriented frame relay and ATM services, along with the connectionless IP and routing facilities.

The new products aim to consolidate the functions of core switches and core routers, while boosting capacity significantly.

The goal is to meet carriers' threefold concerns:

Alcatel, Lucent, Marconi, and Nortel have all confirmed that their next-generation ATM gear – their multiservice core WAN switches, rather – will feature IP routing (see Switch Vendors to Tackle Core Routing).

"By supporting both switching and routing in one platform, we give carriers a choice," says Mike Lisanti, director of product management for the BXR at Marconi. "We don't dictate to them one way or the other. We avoid cell and packet tax. We provide full Layer 3 functionality. We don't just tunnel ATM over IP packets. This gives us greater efficiency and scaleability."

There are many questions surrounding switch strategies. For one thing, MPLS itself is the source of debate (see Poll: Is MPLS BS?, MPLS Spurs Metro Ethernet Debate, Quantum Validates Veritas, and MPLS Gets Lukewarm Reviews). For another, it's highly questionable whether switch vendors really can displace core routers. The router vendors themselves, including Cisco and Juniper Networks Inc. (Nasdaq: JNPR), are hoping their use of MPLS will eliminate the need for ATM at all. (The jury’s out on that one, too.)

Some vendors are choosing to sidestep the routing issue entirely. Équipe Communications, for instance, says its Équipe 3200 (É3200) platform allows carriers to move their ATM networks to MPLS functionality without having to install a separate IP infrastructure first (see Équipe: Take the ATM Road to MPLS).

Équipe is careful to step clear of the routing fray in its marketing as well. "We have never said anything about supporting native IP," insists VP of marketing Bob Sullebarger.

Others are picking spots away from the core. WaveSmith, for instance, has targeted the edge of the carrier network, not the core. "There's a heck of a lot more opportunity for us at the edge," Dunn says. Still, he's not discounting the impact multiservice core switches from key players will have on the future of carrier networks.

It's important to note that, regardless of their product choices, carriers still have a long row to hoe when it comes to migrating ATM or frame relay traffic to packet-based networks. Support of MPLS in a router or switch doesn't mean that traffic will be automatically converted from one to the other.

Instead, a range of tunneling options are being explored in various standards bodies, including the Internet Engineering Task Force (IETF).

What's more, signaling specs, such as GMPLS (generalized MPLS), which may require the basic MPLS protocol set, are being constructed that add yet more choices and complexity for carriers (see Optical Signaling Systems).

Even with standards, though, MPLS won't be a cinch to implement. "There's a confusion about MPLS," one carrier's VP of engineering stated earlier in 2001. "MPLS isn't just one protocol. It's a whole family of protocols."

He points out that while MPLS enables much more reliable and flexible interaction between routers, protocols like Diffserv must be implemented separately on top of basic MPLS in order to create the quality and class of service support most carriers want and need. And that leads to complications.

"Instead of 7,000 tunnels," he notes, "you need 21,000 tunnels, one for each service class."

The issues that remain to be solved regarding ATM and how it fits with IP aren't stopping carriers from seeking the right blend of old and new in their equipment.

"Every big telecom carrier is looking for a next-generation ATM switch or a next-generation IP/MPLS core switch," says Marconi's Lisanti. The carriers, he says, are telling vendors they want to revamp their gear by late 2002 or early 2003 to support existing multiservice networks – supporting data, voice, and video – while offering a migration strategy to MPLS.

Analysts say this represents a key trend. "A year ago, routing vendors (and some switch vendors) believed that there would be no need for an ATM-enabled switch to have a larger switch capacity than 40Gig," writes Christin Flynn, director at Yankee Group in a presentation given this past fall. "Just when you thought it was time to swap out your ATM switch for something more exciting... Here we are in 2001 and starting to see the availability of switches that start at 50Gig and scale up from there."

The next-generation ATM switch even has a new name – multiservice WAN switch. And industry-watchers say it's due to become a growth market again by the end of 2002.

"Revenues were flat for [the multiservice switch] market for 18 months, but we'll see it start accelerating at the end of next year," says Rick Malone, market research director at Vertical Systems Group.

Others agree. According to figures published this past fall by Yankee Group, the worldwide multiservice switch market will grow 56 percent by 2003, from its previous level of $4.7 billion. By 2005, Yankee predicts, it will be a $10.6 billion market.