SAN ANTONIO, Texas Some of the biggest backers of circuit-switched architectures are acknowledging that packetized voice will overtake their analog world soon, but exactly how remains a bone of contention.
Researchers at the IEEE Globecom conference here this past week scouted the paths to packet, as advocates of centralized "soft switch" media gateways squared off against the supporters of the decentralized Session Initiation Protocol.
If the soft-switch architecture is so appealing, why has it failed to take off outside a small group of experimental carriers? Christof Wahl, president of wireline networks at Siemens AG, ticked off the reasons: xDSL has provided some local-loop congestion relief, the recession has squeezed capital expenditures, carriers have proved unable to market new applications, the demise of competitive carriers has taken pressure off the incumbents and residual regulatory issues have stymied rollouts.
Nevertheless, the decoupling of applications, switching and transport are necessary as networks move to primary Internet Protocol traffic, Wahl said. Siemens and its Unisphere affiliate have a lineup of soft-switch products. Wahl stressed that he still has a $3.5 billion circuit-switched business to support but acknowledged its inevitable obsolescence.
Shifting to an architecture in which an open media gateway controller manages traffic flows between gateways and a packet network will let carriers use 80 percent fewer switching elements, reduce operating expenses 40 percent and realize a 10 percent increase in revenue per user, Wahl said. Overall production costs can be reduced two to five times when a carrier shifts to soft switches, he said.
Mike Hluchyj, founder and chief technology officer of Sonus Networks Inc., said the advantage becomes visible when examining equipment footprint. Control of 50,000 calls in a circuit-switched environment requires 40 bays of circuit switches, compared with two 19-inch racks of soft-switch equipment. So far, the soft switch has been used primarily to replace the tandem and toll features of Class 4 circuit switches, he said. The revolution will come as real Class 5 local-office features come to the digital packet world and are augmented beyond Signaling System 7 features because of the richness of the Internet Protocol.
The real "killer app" for soft switching will arrive when the application server for a gateway becomes the voice equivalent of a Web server for data, Hluchyj said.
Wahl said Siemens already is telling customers to prepare for circuit switches' demise, since larger OEMs will halt support for them in the next few years.
The simplest transitional model is an evolutionary one, where a packet interface is brought to a native time-division multiplexed (TDM) switch.
A more nuanced path is a software-reuse strategy, in which circuit-switching software is ported to a commercial hybrid or packet switch and a commercial operating system.
In the most extreme approach, a customer uses a green-fields location to go directly to a soft-switch architecture. That model is the cheapest and simplest, but customers may not find all Class 4 and 5 telephony features available today, Wahl said.
Rather than let the circuit switch die a long, painful death, OEMs and carriers should work together to put it out of its misery, and the Session Initiation Protocol should be the weapon, said Henry Sinnreich, distinguished member of the engineering staff at WorldCom Inc. and a primary author of SIP. Sinnreich called circuit-switched voice an unprofitable beast that is killing the business prospects of incumbent carriers.
Postponing the inevitable
With SIP, more of the intelligence for call control moves down to client devices such as PCs, telephone systems and cellular handsets. The Megaco/Media Gateway Control Protocol suites, by contrast, focus on centralized control, through Unix or Linux servers, of gateways interfacing with the packet-switching fabric. When enterprises buy IP PBX equipment or when carriers buy soft switches, they are merely postponing the inevitable, Sinnreich asserted.
SIP is in good shape to become an end point of choice, he said, particularly since Microsoft Corp. has embedded SIP in the Messenger application in Windows XP. But if SIP is to win, he said, the world must alter its centralized-control, voice-centric-hardware and single-vendor-bundling views of telephony.
"The circuit-switched market is in its last throes," he said. "Better to let the Class 4 and 5 switches simply die."
But Christine Hartman, an analyst at Probe Research, said that the world is a long way from a move to all-packetized voice. Best case, packetized voice will represent only 14 percent of all voice minutes by 2006, she said. TDM services will remain in networks for 20 years or more, she predicted. By 2006, voice will account for 25 percent of the total traffic in the transport core but less than 10 percent of all local traffic, Hartman said.
It is not competition on the wireline side that will force incumbent carriers toward packetized voice, she said, so the demise of CLECs will have little effect on soft switches. Rather,wireless carriers will pose the biggest threat as more customers opt to make their mobile phone their only phone. Wireline carriers must offer advanced digital services to retain voice customers, she said, and that means a turn to packetized voice.
Companies that emphasize multimedia "presence" applications, such as RainDance and Astound, will help drive soft-switch trials, Hartman said. But software vendors in operational-support systems and in operations, administration and maintenance have a lot of catching up to do to make OSS/OAM software as sophisticated for IP as it is for circuit switches.
Nancy Lambros, lead member of the technical staff at the Technology Resources division of SBC Communications, threw cold water on all the presenters by reciting a litany of the problems encountered in soft-switch hardware trials. Lambros put first-generation equipment from Cisco, Lucent and Nortel under scrutiny and found all the vendors wanting in several areas.
The earliest IP PBXes failed because they provided only for PC-based soft phones, she said. If a packet PBX or gateway doesn't have a bundled "hard phone" to offer, the architecture cannot succeed.
Corporate IT managers have not taken to heart what it means to shift the telephony infrastructure to IP, Lambros said. Some elements are obvious. Some are subtle. If a phone uses the Dynamic Host Configuration Protocol, it can be shifted from building to building without a call to the carrier, but corporate IT may have to provision several UDP ports for adequate voice service. Few would be willing, for security reasons, to have open UDP ports for IP telephony.
As for IP quality-of-service, Lambros warned users not to count on it. In theory, the Differentiated Services spec will provide 64 levels of packet priority. But in practice, DiffServ mapped to a Multiprotocol Label Switching topology backs down to four or five priority levels, perhaps not good enough for high-quality voice, she said.
As for the continued squabble among advocates of H.323, Megaco and SIP end points, Lambros warned that protocol wars could be a packet voice killer.
"Some of these issues are not deal killers. Interoperability, though, could be a show stopper," Lambros said. "The last thing you want to do is get your initial trial network stuck on a protocol island, where newer networks around you move to incompatible protocols."