Ethernet will wash ATM right out of the network

Peter Judge: Ethernet is like tap water; ATM is like Evian. Which would you use to wash your car?

The continued fight between Ethernet and ATM (asynchronous transfer mode) has been one of the IT world's longest-running. And one of the easiest to predict. Ethernet just keeps coming out on top. Every time.

The next bastion to fall to Ethernet could be the service provider networks providing back-haul for broadband services. Because, once more, the benefits of ATM will turn out to be illusory or irelevant to what the network needs. ATM may provide quality, but fundamentally, the barrier holding back broadband is cost, not quality.

ATM has a stronghold now in the provider networks used to deliver ADSL services to users. Here's the argument: the ATM network gives quality of service, so that service providers can deliver the kind of services that users will be prepared to pay for with real money. Bandwidth is tight in the first mile (between the user and the telephone exchange) so ATM is what we need to eke it out, and make the best of it. Ethernet would just fling frames of data about, without any guarantees. That has been the accepted wisdom in the DSL rollout so far.

And it's true up to a point. I can certainly vouch for the ability of ATM to deliver quality of service to users, even with limited bandwidth. For a year, I subscribed to the Homechoice service in London, which gives true video on demand over ADSL. The service has about 2Mbps to use, and thanks to BT's VideoStream service, it worked really well for video -- I could pause, fast forward and rewind, with my remote control talking across the Net to a server ten miles away.

I was deeply impressed by the service, but I don't use it any more, simply because I'm a curmudgeon who can't be bothered to consume all the broadcast media that I'm expected to -- even when I can time shift it any way I want. I didn't think 24-hour cartoons for the kids were a great benefit either. And the Internet part of the deal, where I could do what I wanted, was only at a little over 100Kbps.

Something like Homechoice may suit lots of people, but the long-term viability of this kind of thing must be in question. Homechoice subsidises the Videostream service heavily, because users would never pay what it actually costs.

And that is why we need Ethernet instead. What users want is masses of bandwidth. Beyond a certain basic level, we don't really mind about the quality.

ATM is the equivalent of offering us Evian water. We don't want it. We want tap water, to run hot baths with, cook with and wash our cars with. ATM may deliver the best-quality data in the world, but so what? Service providers simply cannot give us enough of it, cheaply enough, to do all we want to do.

If you're outside the coverage of DSL, you may say I'm carping. If only you could get it, you'd probably be perfectly happy to buy the current DSL offer at the current price. You don't want it cheaper, you want it now.

But think about it. Price has a bearing on the basic availability of broadband. If it were cheaper to wire up an exchange for broadband, then BT would need fewer subscribers before it wired your exchange up. Cheaper bandwidth could mean BT setting lower trigger levels.

The price of Gigabit Ethernet ports is tumbling. Once again, Ethernet has commoditised itself and elbowed ATM aside. Ericsson has launched a cheap DSL product based on Ethernet uplinks and others will do the same.

Of course, it will take a while for this to happen. The service providers have invested heavily in an ATM-based rollout (and indeed that was the best approach at the time) and will be expecting to write that off over a period of three years.

But Ethernet is as cheap as tap water. It will be a consumable. Ericsson reckons service providers can stick Ethernet-based kit into telephone exchanges and write it off over a single year. The DSLAM talks to a standard Ethernet switch, so the service provider isn't locked in. That's pretty impressive.

At this stage -- and pretty much for ever -- we need standard DSL versions between the telephone exchange and the user. But plumbing the rest of network with cheap data has got to be the way forward. With cheap Ethernet switches linking the exchange to the data network, contention ratios will cease to worry us. A thousand DSL users won't be sharing one 155Mbps ATM channel, they'll have a Gigabit or two -- and for less money.

I have no doubt that in due course, the cheap Ethernet flood will simply wash ATM out of the network.

I'll drink to that.

To have your say online click on TalkBack and go to the ZDNet UK forums.