Protocol standoff! Version 4 of the H.323 call control protocol for IP telephony is just about ready to hit the streets, but SIP has been stealing the spotlight as an IP-centric apps enabler. Many carriers want both – Online – Session Initiation Protocol
John C. Tanner
Voice over IP has come a long way since its days as a PC hobbyist toy and a neat way to make poor quality but cheap long-distance calls. The quality has improved so much that even international carriers use it to save money on costly backbone routes. International VoIP wholesaler ITXC claims that up to 30% of its IP telephony traffic comes from Tier-1 carriers. TeleGeography estimates that 6% of IDD traffic last year was VoIP.
Behind the scenes, however, there has been an ongoing and frequently passionate battle over the signaling protocols that add call control to IP telephony. On one side is H.323, the global umbrella standard from the International Telecommunication Union. On the other is SIP (Session Initiation Protocol), a signaling protocol adopted in early 1999 by the Internet Engineering Task Force.
Thanks largely to initial deployment and interoperability problems, H.323 has taken something of a PR beating over the years, while IP telephony proponents and Net-heads in general have been increasingly enthusiastic about SIP as the only signaling protocol that makes sense in a world that, one day, will run on nothing but IP. SIP has some strong backing from players like Microsoft, which is including SIP capabilities in the instant messaging application of its new Windows XP operating system. Even the wireless world is behind SIP, with the 3GPP group responsible for UMTS/W-CDMA standards saying it will incorporate SIP into 3G mobile systems.
However, H.323 is far from a lame duck. Version 4 of the standard, which was finalized in November 2000, is ready for commercialization. In late 2001, RADVision, which manufactures the H.323 protocol stacks used by the majority of vendors who make H.323-compliant products, released its Version 4, meaning that Version 4 products are on the way, probably later this year.
How much demand there is for either technology, as always, depends on who you ask, but with each technology possessing its strengths and weaknesses, and with H.323 having a legacy base that SIP does not, some vendors and market analysts see enough room in the next-generation network for both H.323 and SIR
New and improved H.323
The rivalry between H.323 and SIP is a manifestation of the overall rivalry between the circuit-switched telephony world and the (mostly, if not completely) IP-based packet-switched world.
H.323 originated in the circuit-switched telephony world as a group of protocols to allow better managed, carrier-quality IP telephony on enterprise LANs via ATM or ISDN. However, explains Christian Biller, Siemens Information and Communication Networks’ product line manager for multiservice switching, as VoIP gained popularity as a cheap long-distance tool, “many service providers responded to the growing customer demand for cheap Internet calls by deploying H.323 equipment to deliver `carrier-grade’ VoIP with dedicated quality of service”.
H.323 has developed over the years to support an impressive set of telephony features similar to circuit-switched services. Version 4 adds a few more key features, perhaps the most notable of which is IP trunking.
“In a carrier network, you’ll typically have a gatekeeper that sits in front of a bank of gateways, so that when calls are coming in over the VoIP network and they’re outbound for the PSTN, you want to make sure that call terminates correctly,” explains Paul Jones, voice architecture software engineer for Cisco systems, and chairman of ITU Study Group 16, which oversees H.323 standardization, and the editor of the H.323 Version 4 standard. “Version 4 gives the gateway the ability to signal to the gatekeeper exactly how many DSos it has available, and also to be able to do that on a trunk group basis. You can logically group trunks of DSos together, so you might have trunk group A for termination to AT&T’s network, and trunk group B that goes off to MCI’s network.”
Other features in H.323 Version 4 are actually enhancements or better defined specs for existing H.323 features, such as fax/voice switchover, and in-band tones and announcements (such as “The number you have dialed is no longer in service”) that allow the caller to hear the announcement without being billed for a connected call. H.323 Version 4 also better defines gatekeeper redundancy.
H.323 Version 4 also allows for two different types of call-set-up — Fast Connect and H.245 tunneling — to run simultaneously rather than concurrently during the call-set up process.
“That means we can set up an audio call in both directions as before, but also fully exchange terminal capabilities [such as exchanging DTMF digits] with the end points in two messages,” says Jones — “two messages” being one from the caller terminal and one from the receiver terminal. “So you can set up two-way audio and video without any further signaling, which improves performance by cutting down on signaling time.”
Jones says that Version 4 brings “a lot of stability and maturity to the protocol”. Even so, it has been H.323’s history of vaguely defined specs that has prompted criticism of the standard, particularly from SIP proponents who dismiss H.323 as so faulty and needlessly complex that interoperability problems between H.32.3 problems are inevitable.
Jones has heard it all before and dismisses the critics, saying that interoperability is much less of a problem now for H.323. “We do see some interop problems, of course, but they’re usually fairly vendor-specific. It’s not that the standard is wrong, it’s just that the vendor decided to do something a bit strange – they want to cut a corner, or they think, `Oh, this isn’t so important, we’ll just leave it out’.”
Jones observes that SIP also has the same interoperability issues that H.323 has had to deal with, including undefined areas such as resolving addresses and sending DTMF digits, and vendors noodling with the specs.
“If you want to see a nice example of that, you can look at Microsoft XP with SIP built in,” he suggests. “I can guarantee you that Microsoft, because they’re Microsoft, has added things that are not in the spec anywhere.”
Another example Jones offers is the 3GPP’s adoption of SIP. “The truth is that they adopted it as a starting point, but they’ve mutated the protocol into something that’s usable only within their network. Why? Because SIP by itself doesn’t usually meet all the requirements. There’s a lot of things you need to add in the protocol to support the mobile environment, especially in areas of authentication, registration, mobility in general. When you turn on the phone, a SIP register message by itself just doesn’t cut it.”
The rise of SIP
Unlike H.323, which was designed to work with circuit-switched systems, SIP is a pure IP creation, originating within the IETF as one component of the multicast backbone (or Mbone) protocols, and modeled on other well-established Internet standards like HTTP.
More to the point, perhaps, SIP is designed to reuse existing IETF protocol tools, including MIME (multipurpose Internet mail extension), URLs and session description protocol (SDP), so that it integrates easily with other IP apps such as Web browsing and email.
In other words, says Siemens’ Biller, SIP can do a lot more than set up and tear down IP telephony calls, which has been attracting a lot of attention from carders, particularly with packet switching being the future of telephony and softswitches just starting to hit the market.
“Many in the industry say SIPs real value lies in the fact that it makes possible the development of the next-generation internet, an application-driven, converged IP voice-data network,” Biller says.
Biller argues that SIPs Internet orientation gives it three characteristics that are “crucial to next-generation networking requirements: scalability, flexibility and extensibility. In addition, it is easy to implement — SIP is used to identify the end user, not the communications device, while other protocols and applications actually manage the call.”
SIP uses a technique known as “forking”, which involves a session initiation (i.e. a call) to the called party at several locations, such as the home, office, car, and on different devices such as a PC, fixed phone, cell phone. SIP also allows the called party to respond to a call via text, voice, video or email.
All this, combined with the growing number of Web-application programmers, compared to telephony-application programmers, is fueling the rapid development of SIP domains and applications, Biller says.
The latest and greatest
The buzz about SIP, in fact, is loud enough that carriers are talking less about H.323 these days, and the arrival of new and improved H.323 Version 4 isn’t doing much to change that.
“I have not heard a heightened demand from carriers for H.323 version 4 over the previous version,” says Mindy Hiebert, a carrier convergence infrastructure analyst with Yankee Group, who goes on to say that, without exception, carriers are requiring vendors to provide SIP-capable platforms today.
“Carriers state that they will `tolerate’ a migration to get to SIP, but the term is used very loosely,” Hiebert says. “SIP is a critical requirement. Without it, softswitch and applications vendors will find significant obstacles with selling equipment to carriers today, or in the future.”
Meanwhile, she adds: “In the conversations I have had with greenfield carriers, they are deploying SIP only with a `latest and greatest’ attitude regarding what they plan to implement in their networks.”
One probable factor in determining SIP’s adoption rate will be its selling point of easy integration with other IP-based apps at a time when carriers are looking for new revenue-generating enhanced offerings to supplement their main money-making service: voice.
“To position themselves for long-term success, carriers must find ways to combine revenue-generating’ voice services with the IP network’s economies of scale and flexibility,” says Siemens’ Biller. “Certainly H.323-enabled VoIP service is a step in that direction. However, given that there no longer is a significant difference between the price of an IP telephony call and that of a traditional phone call, the initial attraction of VoIP is fading quickly. Service providers therefore recognize they must add value to basic IP telephony service if they are to create sustainable revenue streams, [and] that added value is found in SIP-enabled multimedia applications.”
Yankee Group’s Hiebert confirms that the market has “rallied around SIP as the enabler of packetized voice enhanced services — even though these services are yet to be seen”.
However, none of this is to say that H.323 is on the way out as many SIP proponents claim, or at least hope. In fact, H.323’s main advantage at the moment, Hiebert estimates, is that it’s already here and in widespread use worldwide. “For the carriers using it already, H.323 is `good enough’ for present network initiatives,” Hiebert says.
What that means, essentially, is that there’s more likely to be a long, cozy period of co-existence between H.323 and SIP in the market.
Biller takes this idea a step further: “To service providers and vendors looking beyond the IP telephony market, SIP and H.323 are not competing standards,” he says. “Rather, they regard each as having a legitimate role to play in next-generation networks. Established and new providers alike are seeking IP networking solutions that support both SIP and H.323, especially incumbent operators who must be able to migrate their networks to all-IP infrastures while still protecting their investments in circuit-switched gear.”
Jones agrees on the latter point, offering as an example recent announcements by some US carriers that they have agreed to terminate SIP traffic originating from Windows XP.
“What you’ll have is carriers taking PSTN calls and sending them through the network using H.323 to control call routing to make sure the calls get terminated correctly, and then they’re taking SIP calls from Windows XP and terminating them into the same network, and because their gateways are bilingual and can understand SIP and 323, they can successfully terminate both kinds of calls. They’re leveraging 323 to the benefit of SIP, which is quite interesting.”
Dual-mode H.323/SIP solutions present an interworking functionality (IWF) challenge, a standard for which is currently being worked on by the IETR However, observes Jones: “SIP does some things different than H.323, so something might need to change in one of the standards for that to happen, And since SIP hasn’t seen wide deployment yet, in my mind it would be easier to make the change in SIP.”
SIP proponents may complain, but they may not have a choice if carriers demand H.323/SIP IWF — which they will, assures Yankee Group’s Hiebert.
“SIP/H.323 IWF is a specific need for many carriers today, with several requests for proposals (RFPs) for IWF out from domestic and international carriers,” Hiebert says. “For SIP and H.323 to co-exist, there must be a way for these two protocols to interwork with one another in the network.”
More big decisions
When it comes to decisions on which protocols to back in next-gen IP network, H.323 and SIP are just two of many protocols that service providers have to sift through, notes Yankee Group senior analyst Mindy Hiebert.
“By just naming a few — SIP, SIP-T, SOAP, H.323, Sigtran, TAPI, MGCP, Megaco, JAIN, Parlay, vXML, CPL, BICC — the reason for confusion becomes self-evident,” she says.
MGCP and Megaco/H.248, for example, both handle call control signaling specifically between softswitches and media gateways. SIP, XML, JAIN and Parlay are more concerned with the API (application programming interface) space, where third party software is developed for platforms such as application and feature servers to enable carriers to create new packet-based services that would be harder to implement on less flexible circuit-switched networks.
As service providers rarely if ever make uniform choices regarding competing protocol. some vendors are deciding to back more than one horse. This is already happening wit H.323/SIP, and Hiebert recommends vendors do the same with other convergence protocols.
“With numerous standards providing similar functionality, carriers are leery to choose the `wrong’ one,” Hiebert warns. “Vendors must demonstrate competency in delivering the next killer app and would be wise to consider outsourcing protocol development and incorporation to vendors specializing in such efforts such as Ulticom, ADC NewNet, or Trillium/Intel.”
–John C. Tanner
COPYRIGHT 2002 Advanstar Communications, Inc.
COPYRIGHT 2004 Gale Group