A final column about ATM?

A bunch of us were sitting in a restaurant near Napa, Calif., drinking some nice wine. The group consisted of me, some venture capitalists, and our significant others. At one point in the evening the talk naturally turned to ATM.

One of the venture capitalists said that as far as he was concerned, ATM’s role was only as the access technology for the last 100 feet. That seemed reasonable, but I’m not sure he was quite right. (At this point I expect my editor will want to expand the ATM acronym to “Asynchronous Transfer Mode,” but I think that would be more than a bit silly. It would be one thing if the expansion produced something that made any sense, but quite another when it produces something that sounds like the name of a bad punk rock band.)

To some people ATM is closer to a religion than a technology. You can tell most of the true believers by the slight Bell shape to their heads, but a few have been under cover – able to masquerade as normal Internet geeks. Talking about the future of ATM with true believers, or with the knee-jerk ATM abolitionists, is a waste of time.

Luckily the real world shows up every now and then and renders many absolutist positions irrelevant. After a while it became clear to even the most ardent ATM fan that 155Mbps ATM to the desktop at the same or higher price as gigabit Ethernet was not a good strategic plan.

The venture capitalist might be correct in thinking that a good place for ATM is in access link multiplexing, but he was ignoring the presence of many ATM true believers in the traditional telephone world. Because they cannot conceive of a datagram network that could provide the quality-of-service (QoS) service-level agreements that they think they need, they will continue to use ATM in their networks. People from the datagram world who know that the right architecture in a datagram network will do just as well might have a competitive advantage, but the phone folks have the money these days.

The other place where this venture capitalist might be wrong is the stuff that looks like ATM in the access networks, asymmetric DSL being an example, is not “real” ATM. Rather, it’s just ATM cells – there are none of the QoS features that defined ATM for most people.

This may be my last column about ATM – it’s hard to get too worked up about a technology whose relevance to real-world data networks is as tenuous as ATM’s is (sort of like the relevance of Thunderbird to the wine we had last night).

But then again, some of the ATM folks are now disguised as Multi-protocol Label Switching (MPLS) proponents, so there may be reason to bring ATM up again.

Bradner is a consultant with Harvard University