You knew something strange was going on when Bob Metcalfe, co-founder of 3Com Corporation and one-time resident of Xerox PARC, decided to proclaim the coming triumph of Asynchronous Transfer Mode – ATM to its friends.
You’d think that having helped invent Ethernet in its modern form he’s stick up for it. But, for a while at least, he too fell under the grip of the connection-oriented switching, end-to-end, 53-byte cell voodoo. What he should really have been telling people was to take the ATM switches out of their networks – especially those doing cuckoo in the LAN – line them up at a safe distance in a field and use them for target practice.
To most sane folk, those three letters stood for ‘Automated Teller Machine’ or perhaps ‘Adobe Type Manager’, but to an industry in heat it spelled revolution and, we were later to discover, big pots of money. Except that the real money was not made from selling ATM, but from selling the companies clever enough to make it. Terry Mathews' Newbridge Networks disappeared into Alcatel for an extraordinary $7.1 Billion, while the cultish Pittsburgh outfit Fore Networks (remember ‘Networks of Steel’?) ended up as another massive write down in the Marconi wreck of 2001-2002. Bang went another $4.2 Billion, mostly - and unbelieveably - in the form of cash.
That was then…
Nowadays enterprise ATM is now pretty much a byword for ‘legacy’, that deadly moniker that attaches itself to technologies deemed surplus to history. There are still customers who run it, of course (and carriers still have plenty of it), but its days as a technology organisations seriously consider buying to do anything more than patch or extend their current investment are over for good.
Even the ATM Forum, the industry body tasked to market it, has turned pragmatic. “Switch developments have largely dried up,” admitted Andy Bray, the UK representative of the Forum when asked about the technology’s decline last week. He suggested the focus was now on developing hybrid switch products, capable of integrating ATM with MPLS (multi-protocol LAN Switching)-based IP switching.
At Telindus, the integrator that bought ATM specialist K-Net in 1998, there are no long faces. According to the company’s director of specialist services, James Dunn, where K-Net had once sold ATM aplenty, the technology now accounts for only 5 percent of sales. “It is not a technology market now - it is about services,” was his comment.
Underlying this change of fortune have been a number of factors, including the recession in spending (i.e all technologies are suffering to some extent), and a range of technical arguments long since lost. For the record, it turns out that ATM-like networks (and one might want to add SDH/SONET to the list), have all the features that carriers like them to have but which – keep this to yourself - the users don’t necessarily notice or even need, namely ‘deterministic’ quality of service (QoS).
Ethernet’s big failing in the eyes of those who predicted ATM’s eventual triumph was that it dealt with quality of service hiccups by delaying packet transmission - not a million miles from the way TCP behaves by dropping packets when networks are congested. In point of fact, that’s good because it tells you need to throw more bandwidth at the problem or develop applications that don’t hog so much of it. It tells the enterprise where it has a problem, and lets it see how data is moving (or not moving) around the enterprise.
Token Ring II
Ethernet, the technology once jokingly said to ‘work fine in practice but not on paper’, is now expanding into the metro and campus network space as people revel in its simplicity, relatively low cost and, most of all, ability to run over just about anything including clammy string. Like Token Ring before it, another topology that had multi-level QoS built into it that nobody except IBM cared about, ATM is providing good business for the yard sale enthusiasts. If Ethernet has a way to go in its development, the big winner in all this has been IP.
ATM was an idea thought up by the telecoms industry as a way of making new-fangled data networks conform to the world they had built over the preceding century to serve telephony. That some vendors saw fit to claim that technologies such as Ethernet and TCP/IP were inferior because they didn’t conform to these carefully-drafted schemes tells that the communications industry was struggling to understand the new interaction between business and technology.
ATM had the misfortune to be born into an age when communications and the technology underlying it is becoming ever more commoditised and the balance of power has shifted from producers to consumers and customers. ATM was always about looking at the world from the producer’s point of view, hence its expense, complexity and inflexibility. That is has declined is another way of saying the industry has grown up a bit. It stands as a warning. Woe betide anyone who tries to pull another ATM because those days have gone for good.