Carrier architects and equipment vendors all have good reason to seem a little “on edge” these days. Multiservice edge, or MSE, routers and their promise of converged edge networks were one of equipment vendors' must-have menu items of the past two years, prompting vendors such as Alcatel, Ciena, Tellabs and ECI Telecom to acquire MSE router vendors and their peers to introduce their own products.

More recently, however, as network convergence among North American carriers has taken a backseat to big broadband, those MSE suppliers have focused more intently on the burgeoning market for triple-play networks, whose unique demands may shift the balance of power in the edge router space while also forcing carriers to face fundamental decisions about network architecture.

Among edge router vendors, the growing proliferation of triple-play networks presents particular promise to Davids hoping to challenge market Goliaths Cisco Systems and Juniper Networks. Those challengers left Cisco and Juniper, collectively, with 7% less of the MSE router market in the second quarter than the duo had three months earlier, and triple-play applications may be no small part of the trend.

“Juniper and Cisco have focused so much attention on each other in the core and some areas of the edge, while these non-incumbent [vendors] have come in and rallied around the triple play,” said Eve Griliches, IDC research manager. As edge networks continue to evolve, vendors' standings could become more volatile because edge router life cycles are shortening (traditional 5- to 6-year spans are becoming 3- to 5-year spans, Griliches said), giving vendors ever more chances to continually leapfrog one another's technology.

Triple-play initiatives are giving Alcatel a growing piece of the edge router market, allowing the vendor to exploit both its dominance of the broadband access market and its 2003 acquisition of MSE router vendor TiMetra. Alcatel bumped Redback Networks from its place as the MSE sector's third-largest marketshare holder in the second quarter, thanks to sales of its 7750 service router (the TiMetra product) to major carriers such as SBC Communications and France Telecom. The upset could well reverse itself next quarter (a delay from one Asian customer pushed some Redback revenue ahead a quarter), but Alcatel's advance is no anomaly.

“The incumbent router vendors have taken notice of Alcatel's success, and now they're figuring out how to deal with the larger competitive landscape,” said Mark Bieberich, Yankee Group analyst.

In June, Juniper Networks unveiled its new E320 Broadband Services router, a 100 Gb/s box designed specifically to meet the big per-subscriber bandwidth and scalability demands of triple-play networks along with the high availability to ensure uninterrupted video service. (Now in trials, the E320 becomes generally available Wednesday.) One of Juniper's key strengths — subscriber management — has been a weakness for Alcatel, forcing it to resell the wares of its competitor, Redback (in fact, more than 10% of Redback's revenue last quarter came through Alcatel). Alcatel finally unveiled its own subscriber management device, the 5750, this summer, which allows subscribers to dynamically provision their own bandwidth increases, either temporarily or permanently. But both Alcatel and Juniper will have to prove the worth of their new products.

“The downside to having a new box is that it's a new box,” Bieberich said.

Redback, with its particular focus on the edge of residential broadband networks, has seen particular success as triple-play networks begin to blossom, claiming half of the world's top 20 carriers — from British Telecom to BellSouth — as its customers. The vendor shipped its SmartEdge service gateway to more than 20 new customers in the second quarter, more than doubling the number of converts it won in the first quarter. And while Juniper and Redback both boast about their subscriber management features, Griliches credits Redback's box with higher GigE density than Juniper's new E320, which could give Redback further advantage as triple-play networks scale up.

“Redback is looking very, very good,” Griliches said.

One of the key divisions among equipment architectures at the edge of the triple play is whether traffic is aggregated in a separate Ethernet switch or as part of the edge router device itself (or not at all). Alcatel divides its routing and aggregation functions between two different devices, using a less expensive Ethernet switch to narrow the number of ports entering the more expensive edge router, thereby lowering the cost of big networks. Some vendors combine these functions in one box, however, arguing that one device is both cheaper and simpler to manage than two. Juniper is one example, as is Israeli start-up ECI Telecom, which entered the edge router market this year by acquiring Laurel Networks. Both vendors use Ethernet blades inside the edge router itself to do the low-cost Layer 2 aggregation.

An inexpensive Ethernet aggregation switch is “something we're looking at, both in terms of internal development and partnerships,” said Rafael Francis, director of product management for ECI's data networking division.

Part of the question involves the debate about where in the network intelligence should reside. Intelligence at the edge might yield more quality control but greater expense and complexity. According to Tom Nolle, founder of the CIMI consultancy, many of the current requests for proposals for triple-play edge routers adhere to the DSL Forum's TR-059 architecture, which specified a fairly centralized intelligence, with routing performed upstream of the access networks. But Alcatel, whose chosen architecture puts more IP intelligence in the outside plant to support multicasting, has another view. According to Lindsay Newell, vice president of marketing for Alcatel's IP routing group, “TR-059 is, for all intents and purposes, dead.”

In architecture strategies for triple-play networks, scale becomes a key concern. And although the one-box approach has its advantages, it doesn't match the scale of the dual-edge approach. For instance, Francis won't promise that all the interfaces on the 16-port GigE line card in ECI's ST200 platform will support wire speeds, but he suggests carriers can oversubscribe the network there on the assumption that not all consumers are likely to watch HDTV at once. Plenty of carriers are comfortable with that compromise.

“No one has really developed a router/switch that's both a top-notch switch and everything you'd want in an edge router,” Griliches said. “You end up throwing tons of processing and memory at it, and you don't get anything that's really that great. You're going be better on one end or the other.”

Because of the limitations of scale inherent in the integrated edge architecture, carriers with big networks generally want to divide routing and aggregation functions into separate devices, while carriers with smaller networks — or smaller triple-play projects — favor the single-box approach. For example, although SBC picked Alcatel for its massive fiber-to-the-node initiative, South Dakota's SDN Communications chose ECI's platform. However, as carriers make their equipment and architecture decisions based on a wide range of factors, that stereotype doesn't always hold. Some small carriers opt to split their switches and routers at the edge because of divisions in their own in-house expertise.

“You're going have your routing guys and your switching guys,” Griliches said.

BellSouth is another notable exception to the heuristic that bigger carriers want a dual edge. In February, BellSouth notably chose to roll out triple-play services by connecting Alcatel's IP DSLAMs directly into Redback's SmartEdge routers. The BellSouth deployment has been something of a showcase for Redback, which claims to have since used it to convince other carriers to adopt the same architecture.

In late May, Redback introduced a new set of high-density GigE and 10 GigE line cards for in-box Ethernet aggregation. With two different sizes of products, Redback can offer customers a choice of both single- and dual-edge architectures; carriers that prefer a separate box for Ethernet aggregation can use the SmartEdge 800 router for routing and the smaller 400 strictly for Ethernet aggregation. In the company's last earnings call, Redback already called its early success in the standalone Ethernet aggregation market “warm and exciting.”

Another consideration that factors into these architecture decisions is whether or not carriers want to use the gear to serve both residential broadband and business services. Virtually every vendor will say its gear is perfect for both business and residential use, even the ones that simultaneously claim their gear is optimized for residential broadband. And those who invested in true MSE devices will argue for their versatility.

“If you put a traditional [broadband remote access server] directly behind the DSLAM and don't have that aggregation layer, how do you deliver metro Ethernet services to your business customers?” Newell asked. But ECI's Francis pointed out that small businesses are increasingly interested in DSL as low-cost broadband rather than metro Ethernet.

European incumbent carriers are picking the dual-edge approach to help them converge business and residential networks, Newell said, whereas U.S. carriers, more focused on broadband at the moment, are more content to leave those networks separate for now.

“Over the long term, we see things eventually moving toward an integrated approach, but in the near term, both models will exist,” Francis said.

If so, carriers will have to be cautious as their triple-play networks scale ever higher, Griliches said, recalling the hazards of routers from Springtide (which Lucent Technologies acquired in 2000, canceling the product two years later) and CoSine Communications (which fizzled earlier this year).

“They could do absolutely everything except scale,” she said. “They were sold as these wonderful edge routers, but as soon as they had 100 customers on them, the boxes just ground to a halt.”

“Service providers are telling us there are products in the marketplace today that will serve their needs well for the next couple of years, but the message to vendors is clear,” Bieberich said. “They'll have to continue to push the envelope with respect to scalability.”