Anyone who has been following the emergence of Gigabit Ethernet technology will have noticed a very peculiar occurrence - industry cooperation.
There are no the wars between Hewlett-Packard and the rest of the world over the VG AnyLAN specification. Banished are the discussions as to the place of routing in a network, or the relative merits of Layer 2 and Layer 3 switching.
The development of Gigabit Ethernet has garnered the cooperative effort of otherwise earnest rivals such as 3Com, Bay Networks, Cisco, Intel, and (somewhat belatedly) IBM. (Apologies to a host of other industry heavyweights.)But why the sudden onset of harmony? Networking already has a standards-approved high-speed technology - ATM. Although the Gigabit Ethernet feature set was laid down last month, a specification for Gigabit Ethernet isn't expected to be ratified by the IEEE until mid-1998.
And yet already a number of companies are preparing their marketing leaflets for a plethora of Gigabit Ethernet equipment. For instance, 3Com is set to unveil a lineup that includes a stackable switch, a server adaptor, and uplinks for existing desktop, workgroup, and backbone switches, with products shipping in Q2 at prices only two or three times that of Fast Ethernet.
According to Paul Sherer, vice president of technology development at 3Com, the push for Gigabit Ethernet does not arise from the merits or lack thereof within ATM, but from the evolutionary way in which networks have developed. "The point is that a large set of networks tend to grow incrementally and organically," said Sherer, "so therefore a speed up of the existing technology is a simple way for those networks to grow."
Ridiculously fast Ethernet
Sherer worked closely in the development of the Fast Ethernet specification, and has spent considerable time looking at the requirements of Ethernet networks. He believes that as Ethernet is firmly entrenched in tens of thousands of networks around the world, it is a rare instance when an organisation will tear that network out and replace it with ATM, regardless of the possible performance benefits.
"When you're going to ATM there's a tendency to want to rip the network fabric out and put in the ATM fabric. That makes that decision a much bigger decision," said Sherer. "So for the increasingly faster Ethernet technologies, a lot of their popularity has been gained from the fact that it was not perceived to be such a big decision.
"It's fairly clear that for several years to come Ethernet is going to be the dominant workgroup technology, because ATM has to show a cost benefit great enough to overcome the entrenched Ethernet in those workgroups."
Sherer feels that the place where ATM is making its presence strongly felt in existing networks is at the backbone, where its ability to handle multiple data forms at high speed is ideal. Ironically, this is also the ideal home for gigabit Ethernet.
As for the technologies themselves, Sherer believes the biggest thing wrong with ATM is that it wasn't being deployed 15 years ago, while the biggest thing wrong with Ethernet is that it was. "ATM's problem is there is not a huge entrenched set of expertise and knowledge. With Ethernet, the application space it was designed for is not the more modern application space.
That's what ATM is attempting to address - a broader set of applications."
And it is in the application space that most of the debate over ATM versus Gigabit Ethernet is taking place.
ATM has long been touted as having a superior capacity to deliver time dependent data, such as audio and video. Sherer says much of this debate centres around the core technologies within ATM and Ethernet. "When we start looking at what's the fundamental difference between the two, it's cells (for ATM) versus frames (for Gigabit Ethernet). And in certain applications that gives some inherent advantage to ATM, while in certain applications it gives an inherent advantage to Ethernet."
Sherer says it all comes down to quality of service. "I think the quality of service that you can derive at a particular speed will always be better with ATM, and that comes back to the cell versus frame thing. But the question is, can you deliver the quality of service that the end users require?
"I think the answer is that the quality of service that can be provided for LAN telephony, LAN videoconferencing and so forth over existing LAN technologies is adequate and acceptable to customers. So I don't see that the quality of service towards the periphery of the network is an overwhelming concern.
"Certainly as you get to WAN technologies there are certain advantages there, and I would always say that a given speed you can get ultimately a higher quality of service over ATM than you can over Ethernet. But the question is what quality of service is needed?"
More steam to the engine
He feels that even though ATM may win this battle at the base technology, the enormous entrenched Ethernet market is generating lucrative opportunities for companies who can develop technologies that overcome Ethernet's shortcomings.
Included in this is 3Com's PACE technology, which grants priority to various types of network traffic.
"ATM is really designed to be a multiplex transit from a whole bunch of sources at a very high quality. But where do you do that the most - at the backbone. And for the last 100 metres to the end system it has been shown quite definitively that you can have very good quality of service over Ethernet."