Bellheads vs. 'Netheads
By Dawn Bushaus.
As the public switched telephone network and the Internet begin to converge, standards are going to become increasingly important as a way to ensure interoperability among disparate networks with different topologies. But just as there is no consensus on what the public network of the future should look like, there also is confusion about which standards to use and who should be setting them.
In the telecommunications world, the ITU is the standards body that equipment makers and service providers look to for guidance. On the Internet side of the equation, the Internet Engineering Task Force (IETF) is the prime standards setter. In addition to those two groups, there are many vendor consortiums that work to make specific standards more useful.
With so many groups working on the protocols that will apply to converged networks, there are bound to be problems. Already, conflicts are becoming apparent. Consider the debate over whether to use asynchronous transfer mode (ATM) or IP switching in the core of the network. That pits an ITU standard (ATM) against an emerging IETF standard called multiprotocol label switching (MPLS). The telecom establishment-a.k.a. the Bellheads-are solidly behind ATM. In fact, most of the major carriers are building ATM networks capable of carrying voice, data, and video. One of the biggest draws for the standard is its ability to specify quality of service (QoS). That's imperative for real-time services like voice. IP switching can't do that yet-at least not in a multivendor environment.
Still, many diehard IP proponents-a.k.a. the 'Netheads-say IP switching is the way to go. The biggest 'Nethead complaint about ATM is its transmission overhead. ATM partitions traffic into 53-byte cells, each of which contains a 5-byte header. That translates to an overhead of about 10 percent. But the "cell tax" is even higher, 'Netheads point out, because IP packets don't necessarily map efficiently into ATM's 48-byte data space. For instance, 64-byte Ethernet packets-a common packet size in LAN communications-require two ATM cells for transmission, with the second cell going more than half empty. In this case, the cell tax is close to 40 percent.
'Netheads also argue that once the IETF's MPLS standard is complete later this year, that technology will be far superior to ATM. MPLS will provide for interoperability among various vendors' IP switches and promises to give service providers a way to specify QoS. Another important feature of the standard is its promise to allow for ATM and IP interworking, says Daniel Taylor, senior analyst at market researcher Aberdeen Group Inc. (Boston). "Everyone is counting on it," Taylor says.
The standards debate is spilling out toward the network's edge too. The rapid emergence of the voice over IP industry has highlighted interoperability of IP/public switched network gateways as a potential problem area. Those in the ITU camp are backing H.323 to solve this problem, while the IETF is working on its own session initiation protocol (SIP).
Talk to any vendor of voice gateway products and they'll tell you they're planning to make their gear H.323-compliant if they haven't already. H.323 is an ITU standard originally developed to allow IP telephony client software from different vendors to interoperate. Version 2 of the standard, which was ratified in March, lets IP telephony gateways from different vendors communicate.
While much of the industry has been pushing H.323, the IETF has been working on SIP, which is part of a family of conferencing protocols that includes the real-time transport protocol (RTP), the real-time transport control protocol (RTCP), and the session description protocol (SDP). Both RTP and RTCP have been incorporated in H.323, but SDP and SIP have not. SDP and SIP act together as a signaling protocol in the IP world. SDP describes the capabilities required for participating in a multimedia session, and SIP sets up the session between any two devices on the network.
Critics of H.323 say that standard is too oriented to the old circuit-switching model (it uses a phone number addressing scheme instead of IP addresses) and too expensive to implement (the codecs specified as part of the standard are proprietary and require licensing fees). Engineers at MCI Communications Corp. have been among the most critical of H.323. "H.323 breaks the model of having one network for all services," says Henry Sinnreich, member of the executive staff working on Internet development at MCI. "We already have a phone network, and we already have the Internet. Now we have to build an H.323 network that does everything differently."
Ultimately, the IETF has more than a technology battle on its hands as it tries to sway the public network establishment. "I don't think the IETF will ever carry the weight that the ITU does," says Aberdeen Group's Taylor. The ITU has a long, rich history of specifying standards all the way down to the physical interfaces, Taylor says.The 'Netheads don't have the muddy work boots and helmets," he adds. "They're just dealing with Layer 3 and above."