What's the Future for Voice Interconnection?
Up to this point, voice networks have interconnected with each other primarily using bilateral agreements between major global providers, primarily using the Class 4 switch network and settlement regimes with a high degree of government input on the actual prices charged.
Internet traffic has been exchanged using different principles, primarily using commercial agreements between providers based on settlement-free peering between large networks and payment of transit fees by smaller networks and providers.
As the world moves towards all-IP traffic, issues naturally arise about how the new frameworks will operate. Voice peering federations and fabrics have arisen with the advent of IP-based voice services. But that’s about all market observers can agree on.
Some think the technology change itself will create need for interconnection methods based solely on electronic numbering while others argue there’s no identity between technology platform change and the business rules governing network interconnection.
To wit, a VoIP provider requires some IP-based form of interconnection to interconnect natively, end to end, with other VoIP networks. The alternative is to launder 100 percent of traffic through the public switched telephone network, using media gateways.
So for technical as well as cost and performance reasons, VoIP providers operating natively in the IP domain require some form of IP interconnection or VoIP peering. The same sorts of commercial drivers, though growing, do not necessarily apply to large incumbent service providers or carriers whose traffic remains preponderantly in the time division multiplex realm.
In fact, settlement-free peering, as distinct from some form of interconnection based on today’s international settlement regime, arguably is not in the best financial interests of any large incumbent with lots of voice traffic and revenue in the PSTN realm.
Another hard issue is how voice peering and interconnection fits with the rapidly-changing market for new applications and services of a multimedia type. Specifically, as more innovation is occurring in the area of Web services, how does ENUM and voice peering fit?
And perhaps that’s at least partly why movement to wider use of ENUM-based peering has been slower of late than many observers had predicted would be the case, even as the GSM Association and various cable operator consortia are preparing to launch peering federations of their own.
And though the GSMA and cable federations initially are focused on cost savings and avoidance of payments to carriers, Jim Farmer, Switch & Data senior product manager argues that voice peering has “to move from cost savings to new revenue.”
Mobile peering, at any rate, is a huge growth area, says Ajay Joseph, iBasis VP, in some part because there are so many providers and bilateral agreements are time-consuming and burdensome.
“There is no need for 700-plus bilaterals,” he argues, which theoretically could be the case if one mobile provider wants to interconnect with other mobile providers.
But there are other motivations. End-to-end service level agreements are one example. So are methods for guaranteeing that a transaction took place, with all of the required business rules. Peering might even help providers preserve calling number information.
Likewise, text message hubbing is more manageable than bilateral short message service agreements. Content services are another area where peering might help. IP Multimedia Services also requires charging based on value or types of sessions, not just quality, adds Joseph.
Also, just because carriers and service providers use electronic numbering or peering federations as tools does not necessarily mean the interconnection business models change.
But peering is not simply a matter of concern for carriers. Enterprises also use VoIP, and might benefit from peering. About 50 percent of enterprise sites have IP PBX capability, though they might not use those features, says Tom Kershaw, VeriSign VP.
“For enterprises, voice is a cost, not revenue,” Kershaw says. “It also is a boring cost.” And in the enterprise market, “ENUM is deader than dead,” says Kershaw. “SIP as originally envisioned also is dead.”
“We were totally wrong about how it would work,” he says. “Nobody has any plans to expose end point addresses to anybody else.”
All the enterprises are doing is dumping traffic from a PBX to session border controller and sending it on its way. “There is no SIP in the enterprise because SIP degrades use of PBX features,” he says. As a result, there is not much SIP trunking either.
Of course, SIP can be used to encapsulate traffic. But then SIP is just a transport vehicle, not a services vehicle. “If that is all we are doing, it is a waste of our time,” says Kershaw.
Still, disaster recovery is a genuine need, and VoIP helps there. “Enterprises want backwards compatibility with SS7 and you can’t kill SS7 until you mimic it,” Kershaw says.
“Enterprises want to move numbers seamlessly back and forth between IP and Class 5 networks,” says Kershaw.
Seamus Hourihan, Acme Packet SVP agrees that “you have to mimic the PSTN to replace it.” And there are lots of pain points for IP and unified communications interoperability and peering. Trust remains an issue in part because of distributed denial of service and other attacks, he points out.
Address mediation also is an issue. IPv6 is coming, so networks have to translate, map and auto-discover firewall pin holes, he says. Many signaling protocols and different versions of standards have to be supported.
Codec mediation issues are getting worse, Hourihan points out. “Real-time audio and video have to be supported, and wireless uses different codes as well,” he says.
“We need overload controls, QoS marking and mapping,” Hourihan says. “Failure detection and recovery also are needed.” After an outage, there is a flood of re-registrations that can tax or crash a network.
“Decisions at the edge are getting more complex,” says John Longo, Nextone VP. And it isn’t so clear what “edge” means anymore, says Steve Heap, Arbinet CTO.
“You may want multiple session border controller interconnects, not just one,” Heap says. “Internal routing management is required to do that.”
Transit routing poses other problems. “In some cases, you might want to choose the simpler of several possible routes,” he says. “Then, to add one more level of complexity, you have to integrate with a least cost routing system.”
Carriers might want to select providers and routes based on price, quality and other policies, Heap notes.
Complexity is another issue. Private federations can provide peer discovery and authentication, routing and telephone number lookup and resolution, signaling normalization, settlement and policy enforcement, says Hugh Goldstein, XConnect VP. And then there’s the challenge of video.
Multi-lateral federations can eliminate the sheer complexity of the network relationship issues, Goldstein says.
There are three drivers for federation, says Carlos Dasilva, France Telecom VP. “There is huge pressure on price and increasing volumes.” Routing efficiency also is an issue, as routing complexity is increasing. There also are new product bundles and wideband codecs to support.
Still, federation is about market power and consumer loyalty, says Georges Smine, Nominum senior director. Federation can help prevent erosion of the value provider by the carrier, he says. Federation normally is most attractive at this point to smaller competitors seeking to disintermediate the large carriers.
Someday, all carriers are going to interconnect using some sort of IP fabric. The issue is when, and under what business conditions. “Peering” doesn’t necessarily mean “settlement free” peering.
Consider the stakes. In 2005, 33 percent of U.S. local call revenue was earned for termination. In 2004, the dominant U.S. carriers booked $32 billion in interstate access revenue; $7 billion in switched access and $15 billion in special access. In 2005, 33 percent to 42 percent of U.S. global voice revenue was paid out as termination fees to other carriers.
Those are big enough numbers to convince any rational incumbent voice provider to be very careful indeed about anything that changes the flow of money for interconnection. IP
