Article: The Hard Sell of Softswitches
What happened to the bright future?
americasnetwork.com
December 1, 2001
By: Dan Sweeney
America's Network
Internet telephony always has been remarkable for the extraordinarily high hopes it has aroused among its advocates. And perhaps nowhere have those hopes been more elevated than in the softswitch.
When the product category first appeared four years ago, softswitch manufacturers predicted that their products would quickly supplant the traditional time division multiplexing (TDM) big iron.
Softswitches, with few exceptions, are based on general computing platforms and these platforms are much less expensive than specialized switches at million-dollar-plus prices. Also, a distributed architecture allows a single softswitch to perform the work of several conventional switches at a single location, potentially generating additional savings.
To date, however, relatively few softswitches have been commercially deployed - and those that have are used primarily to address niche applications.
"IP softswitches have been fairly widely employed in Class 4 and Internet offload applications," says Dave Fraley, principal analyst with the Gartner Group. By replacing the Class 4 switches traditionally used in long-distance networks, carriers can save costs by moving long-haul traffic destined for the Internet off more expensive circuit connections.
One such user is Global Crossing. "We're already doing IP transport of voice in the long haul, so softswitches make sense," says John Longo, vice president of data services for the network operator. "We're currently using the Sonus product."
But the largest potential market for softswitches may be as replacements for the Class 5 switches that local carriers have densely deployed worldwide. That, however, is also the market that softswitch vendors have had the greatest difficulty penetrating.
Technical challenges
Three years into its life cycle, the softswitch is either ill-defined or highly differentiated, depending on one's perspective. Besides emulating either Class 4 or Class 5 switches, softswitches can occupy a single box or may consist of up to seven discrete components - termed as, variously, gateways, access gateways, trunking gateways, gate keepers, signaling servers, etc.
A softswitch, by strict definition, uses software to define the switching and signaling functions upon a general purpose computing platform. But at least one company purporting to be a softswitch manufacturer, Taqua, makes its own hardware in seeming contradiction to that statement. The Taqua product also defies convention in providing TDM switching rather than IP, while Oresis makes an ATM and frame relay-based software switch, and Tellica makes a product that supports ATM and frame relay.
GluonNetworks' Avery: Faces ILEC inertia
Still other companies, such as Integral Access, make only the gateway portion of the IP access system and partner with other companies to provide a total solution. All are in a position somewhat analogous to the manufacturers of next generation metro optical equipment in that they have to guess how soon legacy systems will be retired to determine how they will coexist with those systems. And because the softswitches must find their way into variously configured networks, they exhibit a bewildering number of interfaces - all of which adds to development costs and forces softswitch developers to concentrate on specific types of service providers, whether they be ILECs, cable MSOs or DSL firms.
No one is currently offering end-to-end IP phone service over a public network, so at some point the circuit voice stream has to be packetized - an operation that normally occurs in a gateway, which may or may not be housed with the switch itself. A separate gateway may sit at an integrated access device (IAD), a DLC digital loop carrier (DLC), a DSL access multiplexer (DSLAM) or at a residential subscriber premises device. In the case of the new and very expensive IP telephones incorporating session initiation protocol (SIP), no gateway is required because the packetization occurs in the phone itself.
The life of the softswitch designer would be greatly simplified if SIP phones were cheap and ubiquitous, but there's little sign of that happening. "No one on the carrier side is asking us for an all-IP solution," says Rob Avery, vice president of marketing for softswitch start-up GluonNetworks.
"Sure, all-IP is the desideratum," says Terri Griffin, vice president of marketing for Sonus. "But it's a long way off, 20 or 30 years, maybe."
More optimistic is Mike Clement, senior voice engineer for Unisphere. "We can see ahead to a network with no voice switches at all, just routers, a pure IP network. We make a product called the ERX edge routing switch focused on just that kind of future application. It's really not that far off."
But it's certainly not here yet, either."On the local access level, there are no systems currently in place where both gateways and IP switches are present in the same network," says Guy Chenard, vice president of marketing and product development for Integral Access. "We're working with Sonus on some real commercial systems that will be completely IP, but they're not turned on yet."
Because telephone networks are nowhere near to approaching it, the flat, all-IP model has important and unfortunate implications for the softswitch developer. Even the most cursory examination of existing softswitch product reveals the fact that almost all of designers' energies have been devoted to replicating circuit functionality - a necessity for selling products today. And because that functionality includes signaling procedures that are fundamentally alien to the packet environment, circuit emulation over IP involves a much higher level of complexity than does a real old-fashioned circuit connection. Add to that the fact that various access gateways, gatekeepers and trunking gateways have to communicate with one another as well - and note that there are three rival standards for enabling such communications - and the situation of the softswitch manufacturer, particularly in the local access arena, seems almost impossibly difficult.
Despite these challenges, some industry observers say technical concerns have largely been addressed.
"The old concerns about voice quality, scalability, and interoperability have been largely laid to rest," says Tom Valovic, program director for IDC, and softswitch specialist.
Softswitch advocacy
The core arguments of softswitch advocates are two. Softswitches are far cheaper than circuit switches, and they're much more flexible, lending themselves to easier provisioning of individual customers and easier service creation. In most discussions the cost argument tends to take precedence and unquestionably it assumes a certain level of plausibility.
Softswitch vendor Syndeo Corp. has produced a white paper aimed at cable MSOs that provides detailed cost comparisons of the traditional Class 5 switch vs. the softswitch competition showing a near five to one advantage for the latter. Others in the business cite similar figures. But some say these are not apples-to-apples comparisons because softswitches can't do everything that a Class 5 switch can do.
"There are two issues here," says Fraley. "There is the huge feature set of a traditional Class 5 switch with functions numbering in the thousands, and there is the time invested in developing those features and ensuring that they operate reliably. We're talking about millions of lines of code and a million man years of development time by some estimates. A start-up can't duplicate that kind of effort in a couple of years, and that's the justification for the seven figure pricing for the circuit switches."
In many cases, softswitch vendors initially targeted competitive local exchange carriers (CLECs), who did not require the full feature set of traditional switches (Figure 1). " [Softswitches] were mostly designed for the CLEC market," remarks Fraley, adding that only a subset of Class 5 features were built in. "The most competent softswitches might incorporate a few hundred of the thousands of features in a circuit switch. A lot of them only have around thirty, a hundredth of the total set."
But Sonus' Griffin objects to the whole comparison. "If all we're doing is replicating circuit functionality, we're not going to succeed. We have to decide what circuit features to retain and what to replace with new functions. Are things like dial-tone really necessary for voice communications? We think the essential circuit features may be as few as 30."
Ultimately, some say, softswitches will succeed because of their superior flexibility and support for service creation. But if one pores over the advertising literature of the softswitch manufacturers, the argument is not convincing. In almost every case the new service supported by the softswitch turns out to be an existing service already supported by a conventional switches - examples being unified messaging, one number portability, and conferencing.
"I've yet to see a softswitch manufacturer come up with a calling feature that is really new," says Avery. "It's an argument that everyone makes in our business, but it doesn't hold water, at least not now."
"But the potential is there," says Maryling Yu, director of marketing for Syndeo. "An IP softswitch is much better suited to the creation of new services than is the monolithic legacy backplane of a circuit switch."
Here, too, Griffin offers a different take. "I think it's more a matter of ease of use of existing features than creating new ones," says Griffin. "Most people have difficulty remembering the star codes. We provide a more intuitive interface so that more of the calling features become accessible to the caller."
New to you
Now that the CLEC market has dropped off precipitously, softswitch vendors are scrambling for new prospects - especially for their Class 5 replacement offerings (Figure 2). Taqua, Sonus, and Marconi claim to have sold a few switches to rural independents, while Sonus has sold to Time Warner Telecom.
The most notable win was Nortel's sale to Sprint's local division, which plans to replace its Class 5 infrastructure with a softswitch architecture over an eight-year period.
Sprint's key drivers are lower cost to deploy and maintain softswitches and increased revenue streams from being able to quickly provision ADSL, frame relay and ATM, says Mark Chall, Sprint's vice president of network packet switching.
A key component of Sprint's strategy is to also replace all of its digital loop carriers with what Chall calls "line gateways." Such line gateways also will go in every central office, with about one in three central offices also receiving a call server, or softswitch controller. The line gateway will support POTS or DSL service and will convert POTS calls into ATM. (Ultimately, IP may replace ATM as the key transport mechanism.)
Any other local carrier undertaking a softswitch conversion also would have to upgrade its access network, notes John Kuzma, senior analyst for RHK.
"You can't just replace POTS with a softswitch," says Kuzma, noting that unlike Class 5 switches, softswitches do not deliver power to end user devices. "Battery and ground are what's holding back softswitches."
While Sprint is addressing that concern through line gateways, Kuzma does not expect RBOCs to be so willing to make that kind of access network upgrade. Probably the biggest market for in the near term is cable operators. The Packet Cable initiative includes an IP telephony standard, and the softswitch would allow those operators who are still uncommitted to telephony today to forgo the expense of buying a traditional TDM switch.
"We definitely expect to see them," says Michael Harris, president of Kinetic Research, an analyst firm tracking the cable industry. "But concerns persist about the reliability of the softswitches and the suitability of some of the OSS software for cable operations. But what's really holding back cable telephony is the availability of the cable networks themselves. Five nines is not something that has traditionally been required, and it's expensive for the MSOs to get there. When they do, you'll see more telephony and ultimately more softswitches, but that's at least a year away."
Guadelupe Valley Telephone Cooperative, an independent telco based in rural Texas, recently deployed a new softswitch solution from Siemens. The company put an EWSD Class 5 switch in a single central office, interconnecting it with 14 1-Up softswitches - one in each of GVTC's other COs. Initially GVTC doesn't plan to use softswitch functionality, but Robert Hunt, chief operating officer, says the softswitch infrastructure eventually may support packet telephony service over a GVTC owned cable TV network.
Other factors that could influence the popularity of softswitches include the emergence of 4G IP telephony for mobile networks and the maturation of nascent broadband technologies such as free air laser and wireless. Wide deployment of multiprotocol label switching (MPLS) and IP version 6 also could give IP telephony and softswitches a boost.
But widescale success may depend on how quickly incumbent local carriers adopt softswitches - and few are making bullish predictions there.
"We see conventional Class 5 switches being employed overwhelmingly and indefinitely in local access, although we do see softswitches gaining significant market share in other markets such as IP centrex and mobile telephony," says Fraley.
Established carriers' slowness in adopting softswitches is tied to their reluctance to embrace IP telephony, Avery says. "There is a general reluctance on the part of telcos to go over to packets for their mainstream [voice] revenue. You can muster all kinds of arguments based on cost, but they've already bought TDM switches. You're asking them to replace those."
Remarkably, the major vendors of TDM switches themselves, Nortel Networks and Lucent Technologies, are both committed to the softswitch category, although their motivation actually may come from outside the telecom industry.
"You have to have an IP telephony story today," Fraley says, explaining this seeming discrepancy. "The investment community expects it."
One who has such a story is Dan Mangelsdorf, vice president of carrier voice over IP marketing for Nortel. "It is a rather narrow value proposition," he says. "But we can definitely make a case for softswitches in many instances. Copper is the major cap ex for carriers, so in the case of copper exhaust, an IP voice over DSL solution using softswitches makes a lot of sense. You're not replacing the existing infrastructure of TDM switches, you're augmenting it."
Some industry observers still continue to believe that softswitches eventually will fulfill the lofty vision originally associated with them. "It will take time, but the business proposition based on reduced cap ex and op ex will eventually prevail," says Valovic. |
