CROSSING OVER: THE JOURNEY TO PACKETS

After years of pushing voice-over-IP and voice-over-ATM, telecom vendors are suddenly finding little resistance to the idea of packet-based networks. Almost every service provider, from the smallest competitive carrier to the largest IXC, is implementing some kind of circuit-to-packet migration plan. They are no longer asking why. They're asking how.

“Does the technology work? Can it keep up? Those questions are receding rapidly into the background,” said Michael Rubin, director of product management for softswitch vendor Sonus Networks. “We are well beyond worrying about the power and effectiveness of our switches. That's been proved in the field. Migration has become the big issue—getting these switches into the networks.”

Sonus isn't the only vendor seeing a shift in customer thinking. CEOs and CTOs are increasingly resigning themselves to the idea that their networks are evolving to packets, and they've handed the plan off to their operations and engineering staffs.

That's where the fun begins. Engineers and technicians can't easily start ripping out Class 5 switches and dropping in new softswitches or IP gateways. After all, the basic architecture of these networks has stood unaltered for almost 100 years. The average life of a central office is upwards of 40 years. And millions of customers use and depend on those legacy networks. If vendors thought convincing their carrier customers to make the leap was a challenge, coming up with an actual migration plan could prove even more arduous.

“Carriers are always going to be concerned in situations like these — they're changing the DNA of their networks,” said Steve Garcia, director of voice and multimedia marketing for Nortel Networks. “They're starting to become more comfortable, though, as they see what they are able to do themselves and what other carriers are doing. Nobody wants to be the first to try something new and be made an example of, if it proves to be the wrong plan.”

Carriers are going about these migrations in different ways. Some smaller carriers, like BridgeCom, are using softswitches to become facilities-based and wean themselves from incumbent-owned switches. Although some major carriers are migrating whole hog—MCI has committed to going entirely IP by 2005—larger carriers in general are approaching the daunting project of packetizing their networks in stages. Global Crossing started at its core, sending voice over its network of IP submarine cables. Verizon Communications is tackling the tandem portion of its network, simultaneously addressing IP in the Class 4 switching centers of its gigantic local network and optimizing its link to new long-distance business. Other carriers are leaping directly to the most delicate part of the exchange network—the Class 5 switch.

Sprint opted to discard the conventional wisdom of starting in the network core and moving outward. In May, the carrier made the move to packets and began installing softswitches and voice-over-ATM gateways in its central offices—something few incumbents could tackle because of the project's scale and cost. Sprint has the advantage of size, though. While it is one of the largest independent carriers, its total local exchange footprint still encompasses only 5% of the U.S. population, spread across the country. It has a total of 1500 central offices, compared to the thousands upon thousands owned by the RBOCs.

“We have a relatively flat network,” said Mark Chall, Sprint's vice president of packet switching. “We have less than 40 tandems in our entire footprint, and many housed both Class 4 and Class 5 switches. We just weren't seeing the advantages larger operators are seeing in migrating our tandem offices.”

But by making the jump to a packet-based Class 5 network, Sprint will eliminate the need to upgrade its tandem switching centers. By virtue of voice over ATM's “classless” architecture, Sprint's new Nortel softswitches will replace both Class 4 and 5 switches, Chall pointed out.

But Sprint doesn't have an easy road ahead of it. By starting its packet overhaul in its central offices, the carrier isn't just transitioning millions of individual lines to a new switch, it's tinkering with the entire gamut of enhanced services it offers. Services from Centrex to call waiting to conference calling are all in the Class 5 switch, and Sprint isn't taking the cutover lightly, Chall said.

The carrier has already converted seven of its central offices—five remote locations and two hosting centers — to packet, and plans to have 28 online by the end of the year. It may be slow going, but Sprint is taking a very methodical approach to the transition in hopes of isolating any technical, software or logistical problems.

Its first switchover didn't exactly go as smoothly as it hoped. Technicians entering the wrong codes caused trunks to fail, and a software glitch created problems with a customer's PBX system. The glitch was corrected in 24 hours with a patch from Nortel and remains the only technology issue Sprint faced to date.

Meanwhile, Sprint's engineering team is getting coding issues down to a precise science with every switch and gateway installation. “We more than halved our trouble tickets from the first complex conversion to the second—we're fine-tuning our network,” Chall said. “The good news is that most of our customers don't even know we've performed a cutover until we send them an e-mail saying it's complete.”

MCI is fond of pointing out that it has the largest IP backbone world. Now it also has the most aggressive circuit-to-packet migration plan to boast about: all IP, all the time, in just two years. So far, MCI has three Nortel softswitches and IP gateways in play and has started moving its customers on to multiservice platforms.

“We wanted to reduce the number of elements in our network and the number of networks themselves,” said Joe Cook, senior vice president of network planning and engineering for MCI. “We have the largest IP infrastructure in the world, and we want to leverage that infrastructure. It's only natural for us to envision that all our services run over IP.”

MCI anticipates that by the end of this year, 25% of the voice traffic on its network will be packet-routed. Cook wants to raise that number to 90% by 2005. MCI's aggressive timeline puts it far ahead of the “cap and grow” 10- to 15-year deployment strategies other carriers are adopting. Instead of capping spending on legacy equipment and expanding and replacing outmoded circuit switches with packet gear, MCI's two- to three-year migration strategy far outpaces the life expectancy of its circuit network.

Though most carriers aren't taking such an aggressive approach, many are adopting plans that will completely overhaul some aspects of their networks while allowing other parts to function as hybrid systems for the life of their circuit gear.

Global Crossing, for example, is opting to eliminate time division multiplexing from its backbone entirely, allowing it to funnel voice around the world on its IP backbone. But it still will use its circuit equipment in the network core.

“We wanted to have a single leveraged backbone, not a separate TDM network,” said Dennis Morton, Global Crossing's vice president of product development. “But our goal was not to do away with our TDM switches. Fifty percent of our calls go in and out of the same switch, and no call on our network ever makes more than two hops to termination.”

So far, Global Crossing has 25% of its long-distance traffic running over its IP backbone, and plans to move 98% of its long-distance off of its decommissioned circuit switches by 2007. But the carrier is sticking to the cap-and-grow model, deploying IP gateways in new markets and offering end-to-end IP to its customers. So while Global Crossing is not ready to abandon TDM entirely, it is preparing for the ultimate IP migration.

Other carriers have entirely different goals for their packet migrations. Equant is no stranger to converged networks. In 1997 it began deploying voice over frame relay to bring all its traffic onto its IP backbone. However, the limitations of frame relay for video and voice became apparent soon afterwards. Once Equant deployed its MPLS core in 1999, it leapt at the opportunity to deploy application softswitches in its network, said Michael Burrell, Equant's senior product manager.

Cox Communications, on the other hand, has absolutely no problems with the circuit-switched architecture it has in place. The problem is that Cox, as a newer telephony player that bundles voice and data with cable, is maxing out Class 5 switches within a year of their installations. It wanted a cheaper way to expand its telephony offering, so it turned to softswitching, which allowed the operator to simply divide the country into three regions with a central intelligent switch controlling each region. That strategy also gave Cox the added benefit of carrying its own subscriber-to-subscriber long-distance service.

Ultimately there is no cookie-cutter model to migrating to packet-based networks, Rubin of Sonus Networks said. No two plans will be the same, because no carriers want to use packet technology the same way.

“If you're a greenfield operation, then more power to you,” he said. “You can just buy softswitches and deploy a network. But all of these big guys already have networks in place. It's impossible for them to just go out and deploy a packet network. They have to plan these things out carefully because those migration issues are now going to be the determining factor as to when and where packet telephony is rolled out.”