The average Internet surfer probably didn't notice, but an important shot in the battle over the future of the Internet was fired on Wednesday, Oct. 23, when the Cells In Frames Alliance, chaired by Scott Brim of Cornell University, announced the completion of version 1.0 of the Cells in Frames protocol specification.
The Cells in Frames Alliance is a consortium including hardware and software vendors, Internet service providers, user organizations and members of the trade press, along with Cornell. It was formed to develop a system that will allow the use of ATM (Asynchronous Transfer Mode) on existing Ethernet networks. Completion of the specification is a big boost for those who are pushing the ATM protocol suite as the best way to provide reliable audio and video communication over both wide-area and local networks. (A "protocol" is a set of rules for transmitting data.)
It's also a boost for Cornell, which expects to save money by using ATM to unify its voice and data networks, removing many of the costs associated with the existing PBX-based telephone system. Cells in Frames (CIF) technology, mostly developed at Cornell, will allow Cornell to do that without a heavy investment in new hardware. CIF will allow desktop computers to use ATM with their existing Ethernet network cards; the only new hardware needed will be at the hub of each Local Area Network (LAN).
Brim has been working for several months to write the specification, with the guidance of the 100 members of the Cells in Frames Alliance. "Just writing a protocol is easy," said Brim, who is a member of the Advanced Technologies and Planning group in Cornell Information Technologies. "Making it efficient and flexible, leaving room for growth in the future, took seven drafts. We're quite pleased with it."
The specification is a set of rules showing how ATM "cells" should be placed in an Ethernet "frame," and how the user's workstation should communicate with a Cells in Frames hub.
On Ethernet, or any other computer network, data travels as a continuous stream of "bits," each of which is an electrical signal representing either 1 or 0. To specify whose data is whose and where it's supposed to go, computers organize the data they send into short streams of bits variously known as "packets," "cells" or "frames," depending on the system.
Each short stream uses a special pattern of bits to mark its beginning and end, and usually contains other information to tell computers what to do with the data in between. It's like a railroad track with no spaces between the trains: You can tell where one train ends and the next begins by looking for the engine and caboose, and the engineer of each train knows where it's going and what's on board.
ATM uses very short streams called "cells," which contain not only data but also control information that tells network computers what kind of data it is and how important it is. One of the big advantages of ATM, proponents claim, is that it can provide different classes of service. Voice data, for instance, can be given a higher priority than text, so it will be delayed very little en route.
Ethernet uses longer data streams called "frames." CIF packs many ATM cells into each Ethernet frame, along with more control information.
Deciding how ATM information should be arranged in a frame sounds simple, but it isn't, Brim said. Applications that use voice have different needs from those that use video or text, he explained, and the trick was to create a system that satisfied each of those needs while not interfering with any of the others. "Vendors said 'that gets in my way' or 'that's too complicated,'; and so on," Brim reported. "We have vendors doing strictly voice and strictly data, and now they're all satisfied."
Brim, and many others who do networking for a living, hope ATM will become the future standard for local and campus area networks, but not everyone agrees. Other proposals are based on the belief that the currently dominant system, known as the TCP/IP protocol suite, is good enough and that the only need is to provide "more bandwidth" -- that is, new hardware that can carry a lot more data in the same period of time.
Brim doesn't think we can solve all the problems just by building bigger pipes. ATM, he explained, also will offer different classes of service, some of which will have very little delay between when a piece of information is sent and when it is received. This means that interactive voice and video can flow without interruption, while lower-priority traffic like e-mail is shunted aside for a few seconds.
"My concern is voice," he explained. "Extremely low delays make all the difference in the world, and you could say the same thing about interactive video conferencing. When you're talking to someone face-to-face, you get instantaneous feedback when that person raises an eyebrow or makes a noise. There's no such thing as too little delay. We'll never get zero delay, but that's the goal."
ATM also offers flow control and improved routing, he said. These are technical improvements that speed data transfer and conserve bandwidth.
Brim discussed ATM and Cells in Frames at two sessions during the Next Generation Network Conference in Washington, D.C., Nov. 12. Version 1.0 of the Cells in Frames specification is posted on the CIF home page at www.cif.cornell.edu.