Optical/IP Networks

ATM: Over and Out?

Stephen Saunders, US editor, Light Reading It's a funny old world, isn't it? Take ATM. Complicated. Expensive. Doesn't scale well. Smells horrible (Okay, I made that last one up). Yet carriers swear by it -- and at it.

That could be about to change. Don't take my word for it; just ask Dr. Lawrence "Larry" Roberts. He's one of the four Magi who are widely credited with creating the Internet, as we know it. (Extra credit for naming the other three. No? It was Leonard Kleinrock, Vinton Cerf, and Robert Kahn).

Now, Roberts has a brain the size of a small planet. So if he thinks it's all up for ATM, you better pay attention. And that's exactly what he told me just the other day, when we were shooting the breeze about scaleability problems on the public network. "Steve," he said (he always calls me Steve), "it's clear that ATM isn't going to be around long enough to make it worth fixing."

Shocking stuff, eh? And, on the face of it, an outrageous assertion! After all, long-distance carriers continue to use ATM as the foundation for their networks. And some equipment startups -- including Équipe Communications Corp. -- are still developing products based on ATM technology, another sign that life remains in this market.

So why am I even giving space to this ATM antagonism? (Other than the fact that I get paid by the word).

Quite a few reasons, actually. Let's start with price. From a cost perspective, getting rid of ATM is the logical thing to do. Today's carriers typically have to deploy a veritable Dagwood of protocols (IP, over ATM, over Sonet, over DWDM). It makes sense to eliminate the gnarliest and most expensive of these: ATM.

This explains why ATM players are having a tough time of it. Some have had enough, and they're not going to take it any more. Network Equipment Technologies Inc. (NYSE: NWK), for example, recently morph'd itself into a service provisioning vendor after its market cap fell to around $200 million -- less than the value of one of David Huber's big toes.

It may also explain why Ignitus, an ATM startup spun in by Lucent Technologies Inc. (NYSE: LU), has disappeared without a trace.

There's another reason why word of ATM woes should be taken seriously: As Dylan -- Bob, not Thomas -- put it, "The times they are a-changin’." ATM-free services are no longer future tense, theoretical propositions; carriers like Exodus Communications Inc. (Nasdaq: EXDS), Phonoscope Communications Ltd., and Yipes Communications Inc. have already deployed ATM-free zones.

Are they going out on a limb? Yes and no, he equivocated. Admittedly, these packet-based networks are still largely lacking in key areas such as reliability. But then again, ATM isn't keeping pace with the evolution of transmission technology. The ATM community has only just agreed on a standard for 10-gigabit interfaces, for example. And while ATM is okay for integrating voice and data, it's going to be hopeless for handling video, which will probably represent most of the traffic on the public networks of the future.

The reason that more service providers have not followed in the tracks of outfits like Yipes is that the current limitations of packet-based technologies restrict these operators from offering crucial services like voice. But that's changing, too, as technology is developed that for the first time could allow carriers to replace ATM without sacrificing quality of service (QOS).

For instance, in the metro market, equipment startups like Mayan Networks Inc. and Redback Networks Inc. (Nasdaq: RBAK) are working on switches that combine IP with Sonet to deliver voice, video, and data -- sans ATM.

Their counterparts in the core, such as Chiaro Networks (which just received $100 million in a third round of funding -- see Chiaro Gets $100M for "Optical Router" ) and Caspian Networks (which was founded by the great Larry Roberts himself -- see Internet Pioneer Plots IP Revolution), are aiming to do the same thing for long-haul nets, by combining optical technologies with advanced IP implementations.

Standards, too, are helping (for once). Service providers already have the option to replace ATM by using RSVP (resource reservation protocol) to set up MPLS (multiprotocol label switching) circuits across a packet backbone. Now the IETF is working on a new specification, called Multi-Protocol Lambda Switching (MPLambdaS), which could extend the same benefits into the optical domain (see Sycamore Faces "Moment of Truth").

MPLS and MPLambdaS have the potential to dinosaur ATM, although it's still early days. There's a huge irony here, because the more folk like the IETF fiddle around with these protocols, the nearer they get to reinventing good old circuit switching, the technology that's proved so successful at delivering QOS in telephony.

It's all jolly exciting, isn't it? Carriers have put up with ATM for years, but only because they had to. Give them an alternative that’s cheaper and easier to use, and their boot heels will be wanderin’.

Stephen Saunders, US Editor, Light Reading http://www.lightreading.com

Be the first to post a comment regarding this story.
Sign In