Comms Suppliers Living On The Edge

CHICAGO " Debate over how to bring broadband data services from the metropolitan network out to the "edge"-or all the way home to the consumer - was the topic on most minds at last week's Supercomm trade show. Engineers gathering here for the annual conference discussed new prospects for unifying the packet-oriented worlds of Ethernet and Internet Protocol with the time-division multiplexed worlds of Sonet and T1/T3 private lines.

Vendors are attacking the problem with Ethernet, Internet and Sonet technologies.

Ethernet now has become the ubiquitous framing method for the enterprise out to metropolitan networks, and is even being applied in some long-haul networks. Traditional Ethernet switch vendors serving corporate markets are selling environmentally hardened versions of their Layer 2 switches to specialized carriers and Internet service providers concentrating on Ethernet traffic in isolation.

Over time, the switches from vendors such as Riverstone Networks and Extreme Networks have added support for advanced routing protocols and queue-based quality-of-service methods, turning the platforms into Layer 3 switches by default and causing many market analysts to refer to the resultant products as routers.

At Layer 3, these OEMs address the second converging market force: the ubiquity of Internet Protocol (IP), particularly in snaring functions from the mainstay telecom protocol of the 1990s, asynchronous transfer mode. ATM was designed to handle low-latency, high-priority traffic such as voice by creating virtual circuit connections that emulated a true time-division multiplexed (TDM) network. By treating data and isochronous traffic as equally partitioned, 53-byte cells, ATM made it easy to carry a mix of traffic types.ATM, however, never made it to the desktop, a necessity for moving from wide-area to local-area networking. As a result, the experts in IP routing developed new concepts for tagging flows of IP information and managing those flows as if they were ATM virtual circuits.

Multiprotocol label switching, or MPLS, became a core standard for all in the public-networking data world and was even extended into optical transport through the generalized-MPLS standard. MPLS became a mature method of controlling IP flows when the Martini and Kompella extensions were added, allowing easy ways of creating virtual private networks and virtual private LAN services, respectively.

The final market force influencing the battle for the edge resulted from the incumbent local-exchange carriers' all-but-total victory over competitive LECs during the recession. The CLECs had built out networks based on IP and Ethernet alone, but the ILECs needed to justify their existing capital equipment based on Sonet rings. The regional Bells were serious about wanting to move to all-IP networks but needed to do so in a way that preserved legacy TDM circuit-switched systems extending to end users, particularly the fiber-based TDM Sonet rings in metropolitan areas.

'God boxes' While the first generation of edge device OEMs-coming out with products in 2000-2001-looked for clever ways to meld resilient packet rings, Sonet and Ethernet, the next crop must assume a long life for Sonet rings currently in the ground. The newest Sonet add-drop multiplexer (ADM) "God boxes" are cost-effective small systems that can speak Ethernet and MPLS. White Rock Networks Inc. set the standard for low-end Sonet access systems, and now everyone from Fujitsu Network Systems to Turin Networks is looking for a "right-sized" ADM.

What are the common assumptions in this sector? Ethernet at 1 Gbit/second and slower will migrate out from enterprises to subtended ring networks, and 10-Gbit Ethernet may be spoken natively all the way to the MAN/WAN core. MPLS and G-MPLS, initiated in the core, will push out to the edge and may intrude upon multisite enterprise networks as well. All traffic will ride on IP at Layer 3, with only small amounts of circuit-switched voice remaining in any global networks by late in the decade.With the advent of Sonet framer chips that support the link capacity adjustment scheme (LCAS), virtual concatenation and the generic framing procedure (GFP), Sonet designs become simpler for the market at large. The recent arrival of ADM-on-chip designs from the likes of Agere Systems, Galazar Networks, Parama Networks and PMC-Sierra have made it feasible to stuff ADM functions and true Sonet termination into the confines of a 1U "pizza box."

Second time around The brave new world of edge aggregation sparked a revolution of sorts at Supercomm. Market analysts were toasting startups that offered systems for the "multiservice edge," while extolling Nortel Networks Inc. (Kanata, Ontario) for defining a category between public and private networks that was something more than an edge router or a glorified Ethernet switch.

Pioneers trolling the edge included Mangrove Systems Inc. (Wallingford, Conn.), which launched an edge-to-central-office system at Supercomm based on what the company calls MetroMPLS, and Hatteras Networks Inc. (Research Triangle Park, N.C.) and World Wide Packets Inc. (Spokane, Wash.), which demonstrated Ethernet access systems for carriers that could coexist with Sonet and MPLS systems in the metropolitan core.

But there was something all too familiar about the accolades to the edge. Three years ago, as the recession hit the local carriers with full force, a new breed of Sonet ADM was promising to be Ethernet-friendly. Meanwhile, the IEEE's 802 LAN/WAN group had defined a new metro topology, resilient packet ring (RPR), that was touted as the way to bring Sonet redundancy to a simpler, packet-optimized world.

Interest in 802.17 RPR remains strong in certain Asian markets, and the RPR Alliance was in Chicago last week touting the benefits of the technology as a transition point between Ethernet in the enterprise and Sonet in the metro core.Startups from 2001 that had promised to work that multiservice edge, though, were absent. Appian Communications Inc. folded in January after new funding from France Telecom failed to come through. Almost simultaneously, Coriolis Networks Inc. ran out of money and closed its doors, despite having garnered Asian carriers' interest in its optical packet access systems. Lantern Communications Inc. was acquired by C-Cor.Net Corp. (State College, Pa.) in April, leaving Luminous Networks Inc. (Cupertino, Calif.) and Corrigent Systems Inc. (San Jose, Calif.) as the only 2000-01 startups left in the RPR Alliance.

Meanwhile, many vendors are still reinventing Sonet. Turin Networks Inc. (Petaluma, Calif.) used Supercomm to launch TraverseEdge, an expansion of its Traverse central-office Sonet box designed to bring in Ethernet traffic. Ciena Corp. (Linthicum, Md.) has added data-aware Sonet features such as GFP and LCAS to its popular CoreDirector digital switching system.

Even as Sonet gets more data-friendly, edge routers are ratcheting up in complexity. Vendors noted as core router players, including Juniper and Caspian, now are combining fixed IP functionality with more complex service multiplexing to play at the edge of the public network. Alcatel recently made a bid for the edge, launching the packet-prioritization architecture acquired last summer from TiMetra Networks Inc. Nortel went the furthest philosophically, however, combining the open GateD software from NextHop with fault-tolerant features to produce the Multiprovider Edge family-not a router at all, Nortel insisted, but a form of route-enabled aggregator.

Are vendors setting themselves up for another fall? Mangrove CEO Jonathan Reeves said two factors have changed since systems from failed edge OEMs like Appian and Coriolis were launched. "We can't underestimate the impact of new data-encapsulation features to increase the utility of Sonet, such as GFP, LCAS and virtual concatenation," said Reeves, who founded and sold optical/ATM companies Sahara Networks Inc. and Sirocco Systems Inc. in the late 1990s. "The standardization and wide availability of these features make all the difference in the world in bringing together Sonet, Ethernet and MPLS."

The other issue, he said, is timing. "With a few very limited exceptions in Asia, service providers just weren't spending money between 2000 and 2003, and some companies with very good architectures just ran out of time. We'll never see the huge boom in carrier spending we saw in the late 1990s, but at least local and regional service providers are purchasing equipment again," said Reeves.While general agreement has been reached on retaining Ethernet, Sonet and IP/MPLS everywhere in the network, there is still a plethora of potential architectures for multiplexing traffic between the edge and metro core. Corporate LAN managers tend to prefer Ethernet-oriented architectures that add support for Sonet and MPLS, because they are most familiar with Ethernet protocols and with traffic patterns centered on best-effort packet data. However, Kevin Wade, marketing director at Turin, said that corporate network managers who have had to lease T1/T3 lines from carriers, or who subscribe to frame relay services, recognize the value in preserving TDM capabilities at the network edge.

Incumbent carriers, on the other hand, presuppose the existence of Sonet. Though many are learning the value of MPLS at the network core, traditional carriers treat Ethernet and MPLS as features to augment existing TDM Sonet backbones. Depending on the vagaries of the framing and encapsulation chip architectures selected by the OEMs, the aggregation boxes can make subtle distinctions in how traffic is combined for delivery to the core.

Fiber-based transport is assumed for the metro ring and, particularly, for supermetro regions, where speeds of 10 and 40 Gbits/s dominate. But it is not enough to assume that raw bandwidth can be thrown at traffic-prioritization problems. Hatteras CEO Kevin Sheehan pointed out that carriers of all types want to transport data, voice and video, which means carrying multiple streams of multimegabit services to compete with existing analog and digital cable TV.

Also, many neighborhoods offer no optical connections, either passive or active, to individual buildings. Therefore, traffic aggregators must assume copper transport in the final mile to the end user. Hatteras' Ethernet systems thus are optimized for copper transport.

The ultimate problem, though, is that business conditions for startups are only marginally better than they were in 2000-01. John Metz, president of market research firm Metz International, noted that more than $1.2 billion in venture money went into interesting startups with unique edge architectures between 1998 and 2000, and scores of these startups went out of business. Carriers may be buying more equipment than they did during the years when everything went dark, he said, but the safe purchases involve Fortune 1000 companies. "Even new technology that is recognized as superior . . . is not enough to win, or sustain a business long term."Many of the 2004 newcomers have architectures that deserve to see the light of day, but Metz said he fears "it ain't a-gonna happen.' Carriers, he said, "are less risk-averse than before. But a startup better have either adequate funding or a partner who will distribute products and potentially acquire [it]."