Startup WaveSmith Networks Inc. hopes to give heavyweight telecom equipment manufacturers something to chew on next Monday when it announces details of its first products: the DN4100 and DN2100.

The boxes aim to help incumbent carriers address a big problem: how to deal with a capacity crisis in their existing ATM (asynchronous transfer mode) networks without buying gear that might become obsolete when -- and if -- MPLS (multiprotocol label switching) becomes a viable alternative.

WaveSmith’s solution looks impressive on paper. Its boxes pack a huge amount of switching capacity into a tiny space and give carriers a lot of flexibility in deciding when and where to shift services onto MPLS infrastructure.

WaveSmith also claims that its software will avoid a recurrence of recent outages on ATM backbones (see In Qwest Outage, ATM Takes Some Heat) and will tie in seamlessly with existing signaling and management systems.

It all sounds too good to be true, according to Geoff Bennett, director of technology at Marconi Communications PLC (Nasdaq/London: MONI), an established manufacturer of this type of gear. Bennett finds it hard to believe that a small startup like Wavesmith could have done so much in so little time with limited resources .

The proof of the pudding will be in the eating. Three carriers -- among them Comcast Corp. (Nasdaq: CMCSA, CMCSK) and Genuity Inc. (Nasdaq: GENU) -- will start testing WaveSmith’s gear in their labs at the end of May, according to Chad Dunn, WaveSmith’s director of product management. First customer shipments are scheduled for the fourth quarter of this year.

Capacity Crisis

One thing's for sure. Regional Bell operating companies (RBOCs) and other incumbent carriers are in desperate need of equipment like WaveSmith's.

Right now, most of them rely on ATM networks to carry money-spinning services such as leased lines, frame relay, and telephony. A significant proportion of them are using Cascade’s switch, the CBX500 from Lucent Technologies Inc. (NYSE: LU), to handle this traffic (see WaveSmith Targets Cascade Switch).

The CBX500 is beginning to show its age. A particular problem is that it can’t support sufficient virtual circuits to keep pace with the rollout of DSL (digital subscriber line) connections, which typically require a couple of ATM virtual circuits apiece. As a result, carriers need to upgrade their switches and are looking for something that’ll work in an ATM or MPLS environment so they can shift from one to the other at their own pace.

Enter WaveSmith. It’s developed a couple of chassis and a set of modules that address these requirements. The DN4100 accommodates five modules and is 7 inches high, while the DN2100 accommodates three modules and is 3.5 inches high. Each has a switching capacity of 30 Gbit/s.

Three of WaveSmith’s modules are used for handling different types of traffic -- ATM, frame relay, and IP/MPLS. Each one can be equipped with a single OC48c (2.5 Gbit/s) port, four OC12 (622 Mbit/s) ports, 16 OC3 (155 Mbit/s) ports, or 32 DS3/E3 (45/34 Mbit/s) ports. As the chassis backplane is protocol agnostic, carriers can shift from ATM to IP/MPLS by simply swapping modules.

A fourth module enables multiple chassis to be linked together using optics so that they operate as one large switch. Eventually, WaveSmith expects to offer aggregate switching capacities of up to 320 Gbit/s. Right now, however, only five chassis can be linked together. “The sweet spot is 40 to 150 Gbit/s,” says Dunn.

High-speed serial connections are used, so the chassis can be separated by distances of up to 50 feet. In this way, they can be tucked into small spaces in different racks that might otherwise be wasted.

WaveSmith’s boxes pack a big punch. A fully populated DN4100 is one fifth the size of Lucent’s CBX500, uses one tenth the power, and has six times the switching capacity, according to Dunn.

A single module can handle 128,000 virtual connections, and a complete assembly can handle 1.92 million. Equally important, each module can set up and tear down 5,000 calls a second. This ensures rapid rerouting of traffic around failures, Dunn adds.

Marconi says it can match this performance, but only because it’s spent many man-decades developing and refining its own ATM software stack. Startups like WaveSmith would almost certainly buy an off-the-shelf ATM stack rather than develop their own -- and such stacks couldn’t deliver anything like these figures, according to Bennett.

WaveSmith’s other claim to fame is that it’s bringing telephony-grade reliability to data networks by adopting a “microkernel” software architecture. In essence, processes don’t share memory, which reduces the risk of problems propagating and causing widespread outages.

The architecture also enables carriers to upgrade software without interrupting traffic and to revert to older software versions if they encounter problems. The inability to do this with Lucent’s software led to the recent ATM outages experienced by AT&T Corp. (NYSE: T) and Qwest Communications International Corp. (NYSE: Q), according to WaveSmith. Lucent was unable to field anybody to comment on this.

WaveSmith says it's got yet another ace up its sleeve: a way of supporting multiple call control methods simultaneously. This "Open Call Model” hasn't been available in data switching gear before, according to Dunn. It enables engineers dealing with different services, such as voice and frame relay, to set up connections using the processes they're familiar with.

The same architecture also allows carriers to control the DN4100 and DN2100 using existing operational system support (OSS) tools. The company plans to secure Osmine compliance certification next year, once it’s got purchase commitments from carriers (see Telcordia's Osmine Gold Mine).

The price of an entry-level WaveSmith switch is $28,000.

— Peter Heywood, Founding Editor, Light Reading
http://www.lightreading.com