Getting Real With Packet Telephony

While there's no doubt the telecommunications industry has embraced packet telephony as the wave of the future, there is more confusion than ever over how to get there.

Where, a year ago, strategists at long distance and new data-based carriers seeking to get into the local voice market confidently predicted they'd do so on the strength of the emerging H.323 IP (Internet Protocol) voice platform, today they're going off in widely disparate directions, anxious to exploit the packet opportunity but finding it hard to agree on a universal approach. The resulting proliferation of strategies, which often vary within a single company depending on what its local network situation is in any given area, has put a premium on the supply of flexible, scalable platforms, much to the dismay of those who were counting on a single approach like H.323.

There really isn't a workable standard at this point, asserts David Nagel, CTO at AT&T and president of AT&T Laboratories. "H.323 doesn't equal H.323 in different vendor implementations, and the OSS (operations support systems) are even more out of synch," he says.

Nor is it likely that ongoing work on refining the set of protocols making up H.323 toward what some people call "Version 3" is going to simplify matters anytime soon, say those who are working on the standard. "There are a bunch of different models to IP telephony that lead to modifying H.323 in different ways, so there's a lot of push and pull in different directions as we move to the next level," says James Toga, director of systems architecture for MetaTel, Inc. and a leader in the H.323 process. "The architecture you choose has a lot to do with where your asset value resides."

Some carriers, like Level 3 Communications, are forging ahead into local as well as long-distance IP telephony. "IP voice works today," says Level 3 CEO James Crowe. "It's not only less expensive (than circuit-switched systems), but we have an opportunity to see prices drop faster than we've ever seen."

The architecture Level 3 is deploying, which is based on call agent control of distributed intelligence via MGCP (Multimedia Gateway Control Protocol), may or may not win out in the standard battles, Crowe adds, but the company stands to lose more by waiting to get started than it does by having to adjust to becoming standards-compliant later on.

Some carriers, like Sprint, believe that no existing IP platform is sufficient to their scaling but want to proceed with packet-based communications anyway, and so are shifting to an ATM- (asynchronous transfer mode) -based approach, where the layer 3 components of TCP/IP are handled instead by the Adaptation Level 2 layer of ATM. "We can make IP voice technology work by over-engineering the network to avoid blocking and provide a predictable level of quality of service for a limited number of customers, but we can't sustain the required performance with that approach as the customer base grows," says Sprint CTO Marty Kaplan. "That's why we have to move to AAL2."

And then there are the carriers like AT&T which, despite their commitment to packet telecommunications, prefer to wait out the evolution a bit longer but want to make sure they can readily migrate to the IP format from circuit-based approaches they're taking today. "We'd love to take advantage of an existing standards-based, off-the-shelf system, but we don't think that's the best approach," says Chuck Kalmanek, a senior engineer at AT&T Labs.

As complex as all this might seem from the perspective of a vendor who wants to offer solutions across all these migration paths, Nortel Networks has come up with a a flexible strategy for carrier migration which it believes covers all the bases, including those in the heavily populated gray area where any one carrier may be looking for multiple paths. "We support a wide range of strategies," says Rod McPherson, vice president, next-generation networks at Nortel Networks. "Many carriers require multiple network solution options because they may be an incumbent in one segment of their operations while they're pursuing a greenfield expansion in another segment."

Nortel Networks' solution strategy for existing service providers is meant to allow these carriers to begin exploiting the feature and service expansion possibilities of packet-based technology while taking advantage of all the assets of their existing multivendor switching infrastructure. Here the bearer path for the packet content is established via the ATM Adaptation Level 1 SVC (switched virtual circuit) while the signaling channel is designed to interface the IP applications with the SS7 (signaling system 7) domain of the switch (from Nortel Networks or another vendor), thereby allowing carriers to exploit the advantages of IP-based applications while accessing the base of over 4,000 services that reside in today's DMS and non-Nortel Networks switching environment.

The first iteration of Nortel Networks' pure IP solution is designed to replicate the basic telephony and other features, such as credit card and local number portability, that cover about 70 to 90 percent of all the calls made today over the PSTN, Bender says. "It's a leaner, less expensive platform (than recreating the PSTN and SS7), but you can put most of the traffic you'd run on the PSTN on our IP solution," he says.

The key to maximizing carrier flexibility in the migration to packet-based telephony is that features common to both an ATM-based and pure IP domain can be passed between them via the interconnection of call servers.

Greater Interoperability

By next February, MGCP should be standardized under the name H.248, leading to greater interoperability among the vendor platforms that are being built to accommodate this increasingly popular architecture, Bender says. But, he adds, this level of interoperability won't extend to the INM level of integrated management, so that carriers will still have to use single vendor sources within any one operating cloud for some time to come or else separately manage the different pieces from different vendors.

Moreover, he adds, as new applications are invented, it may be that the standardized version of MGCP might not be able to handle them, in which case carriers would have to use proprietary adjuncts to the standard. This is why development of a flexible architecture that maintains the tie to the legacy network while accommodating the full range of packet possibilities makes the most sense, he says. Standards will come, but time to market pressures mean customers need to choose partners and suppliers that can get a full range of service out there first.

"People are starting to realize the full range of challenges and opportunities Internet Telephony brings. It is a fundamental shift in the technology we use to communicate, and the implications are far-reaching," Bender says. "The Voice over IP industry we've been living with over the last three years started us in the right direction, but that technology has taken us as far as it can. Internet Telephony is not about bigger gateways and `carrier class.' It's about a radical new architecture which is cheaper, more efficient, more service-rich and far more dynamic, flexible and open than traditional telephony. A lot of people are still asking the wrong questions, but more and more are coming to grips with the pervasiveness of change that Internet Telephony is bringing. Those people are the ones who will invent the network of the future," Bender concludes.