Even in the world of hype that is the computer business, few technologies have been greeted with the enthusiasm that met the announcement of ATM at the beginning of the decade. And few technologies have disappointed more in the intervening period.
ATM is the networking technology that everyone wants but no one wants to buy - although that slowly seems to be changing. Its advantages over sad Ethernet are self-evident for a high-bandwidth world, but it has failed to compete in two important areas: standards, and price. Ethernet is cheap, standard and for high-bandwidth applications, performs like a dog. ATM is expensive, it has lacked standards, and unfortunately, its performance still disappoints - though compared to 10Mbit Ethernet, the disappointment is not that bad.
If ATM is to finally fulfil its potential, the battle lines are being drawn in two areas: first, 155Mbit ATM backbones, where the rival technology is FDDI and the demand is for the network to fly without having to throw away the entire Lan infrastructure.
The second is low-cost ATM to the desktop - a contradiction in terms according to many in the industry. In this market, ATM features much stiffer resistance from evolutions of existing technologies.
Nevertheless, IBM's research shows one in seven UK companies plan to implement IBM over the next 12 to 18 months, compared with one per cent of the sample who have it already. And those potential users are concentrating on the backbone and the Wan, not the desktop connection.
The advantages that many people thought ATM would bring, such as networked multimedia, full-screen streaming video and the convergence of voice and data are almost without exception being ignored. Instead, users are backing ATM because it gives them data bandwidth.
This doesn't mean that ATM is a commodity market for resellers. If customers are not looking at multimedia or voice over ATM today, they won't want to upgrade the network again when they are ready to take the plunge. And achieving interoperability between ATM products from rival vendors is still a minefield for unprepared Vars.
3Com is one major network product supplier that is concentrating solely on the backbone for ATM at the moment. Product marketing manager, ATM and wide area products, Joe Frost, believes that users are not demanding desktop connection yet. "ATM has been considered the panacea for everything, but we concentrate on users who want to build a data network without ripping out everything that they have today. We say, leave what's on the desktop alone and build an ATM backbone. In the majority of these installations, ATM exceeds their requests, and the beauty is that ATM is almost infinitely scaleable," he explained.
3Com has been criticised for being slow in bringing products to market for ATM, but that's due to a lack of demand and a lack of standards, Frost added: "We do have cards that take 25Mbit ATM to the desktop, but the majority of installation will be either 10Mbit or 100Mbit fast Ethernet.
3Com will only bring out standards-compliant products: Lan emulation was finally ratified in February 1995, and we were the first to bring product to market in July. We are not in the proprietary market though."
Another company concentrating on the backbone is reseller K-Net. ATM product manager Andrew Rowney doesn't think that ATM will not succeed on the desktop - but he is yet to see demand coming through. He blames companies like 3Com for holding the market back. "The first market is that for just expanding the network as it stands - powerful PCs can simply flood an existing network. This used to be a problem only for workstation users - Sun or HP clients on a Lan. Now it is a problem for Pentium users too, and FDDI or Fast Ethernet basically have no scaleability. The fear, uncertainty and doubt over ATM is all put about by established companies who cannot supply product, and that's many of the major companies. Fore Systems, as a major supplier that has been providing these ATM products for years, have been fighting this attitude too."
K-Net knows that it can not justify ATM to the desktop in the same way.
"FDDI is a dead technology - anyone putting FDDI in has to have a very good reason for doing it. In our opinion, the battle for the backbone has been won. The battle ATM has to fight instead is for the desktop with its 25Mbit adapters. The price is competitive with Fast Ethernet, but is still expensive. We say, consider leaving your existing Ethernet where it is, and think about using switched Ethernet. Where you have more powerful users, put them directly onto 10Mbit segments," Rowney explained.
This is not IBM's point of view - not surprisingly, as IBM drove the 25Mbit ATM standard through the standards process. Its reasoning was that the cabling that many IBM customers used would not support 155Mbit to the desktop; its tactics were as far from those of 3Com as you could get - producing the product, then agreeing the standard.
Now IBM's commitment to 25Mbit seems to be bearing fruit. Dataquest reports a ten-fold increase of shipments for 25Mbit cards - although with only 11,000 shipped last year, this is not a landslide yet.
Andy Greaves, IBM UK director of network systems, defends this tactic.
"Standards are always difficult - without doubt, ATM has to be open and standards-compliant. But if everyone sits back and waits for standards to be agreed, it acts as a brake on technology. We will produce products ahead of standards being agreed, but if the eventual standard isn't ours, we will change," he said, citing IBM's U-turn on Lan emulation technology as evidence that it was not too proud to reconfigure.
IBM doesn't see 25Mbit as just a convenient technology either. "25Mbit is excellent technology because it is low-cost, and you can run it using existing cabling. You would have to have a pretty hairy application to need 155Mbit at the desktop. The price per desk for our solution now is u350 - and that includes the adaptor, plus your share of the hub. Most people's constraints are not about multimedia, but they are a clogged-up server," Greaves added, admitting that "The market for ATM is still very small."
Digital agrees that the market is small, but also rubbishes IBM's claim that it is only a matter of time until 25Mbit becomes a de facto standard for desktop networking. "ATM still holds the promise of integrating voice, video and data over a WAN. Other technologies don't hold that promise," said Lee Knock, ATM marketing manager for Digital Networks Europe. "But it will be hard to displace personal Ethernet, where a good adapter costs u60. ATM prices will not be in that range for at least three years."
Digital, like IBM, will not put all its efforts into ATM, because many users simply do not want it. "We continue to invest in ATM but we also provide products using IP switching. There has been a definite ramping up of demand for ATM, but not at the rate we expected a few years ago," said Knoch.
He sees a basic set of uses for ATM, and once again, no pull to provide the multimedia functions to take advantage of switching. "The applications we see it used for are basically server consolidation - you put a fat ATM pipe into that server. But for the client, you use switched Ethernet, and either IP switching or Lan emulation. If you made me commit, I would say that the future is cloudy for 25Mbit ATM at the desktop. When you add all the specifications for ATM that are still being decided, it will drive the cost of silicon up. Second, the justification for 25Mbit ATM was that the cabling that IBM users had did not handle 155Mbit, but much of that cabling will be replaced soon. Third, Ethernet with enhancements will still prove cheaper. Fourth, fast Ethernet will continue to drop in price just as ATM does."
He also sees two larger trends that make selling ATM a problem for Vars: a lack of a common songsheet for vendors to sing from, and the lack of software to take advantage of ATM. "Digital's ATM story is different from IBM's, Cisco's, or Fore's and that is hurting the industry. The ATM Forum is trying to do the right thing and establish these standards properly - it is just that all the vendors want their opinions reflected in what it decides. As the vendors thrash the issues out, customers get worried," he said. It would not be so big a problem if ATM was delivering the performance it promises, but its youth and a lack of proven drivers and software are serious impediments. "With TCP/IP on an alpha server, we have achieved a throughput of 134Mbit, and that's the theoretical maximum. I know some PCs with a 155Mbit connection that get a rate of 30Mbits. Even with our NT drivers, a Windows PC gets about 80Mbits."
Mass-market ATM is a goal for SMC as well as IBM - Ian Palmer, systems network engineer at SMC has plans to drive entry prices down, but regrets the lack of software. "The demand for services offered by ATM, such as realtime voice and data, are low. The applications that people have on their desktop don't have those facilities yet, and there are no off-the-shelf ATM applications," he said.
"ATM has been very Unix-orientated, but SMC has been trying to develop new markets with NT and NetWare adaptors, and we find that NT performs as well, if not better than Unix. This will bring ATM to the operating systems that the rest of the world use."
Mass-market does not mean that prices will sink to Ethernet levels though.
"We are aiming to provide cost-effective adaptors - for example if you are putting an adaptor in a server, then ATM requires quite a lot of memory, and we provide 2Mb. But if you are putting it in a client, you need less, so the card can have 512K and be cheaper as a result. So our client ATM card sells for u729," he explained, 'It will be a while before Ethernet price levels are achieved. While standards are still being decided, a lot of the standards are pushed towards a software implementation, but when the standards are in place in a couple of years they will be put into the silicon and be cheaper."
If the current state of ATM represents a judicious trimming of its cure-all claims, there's always the promise of business for resellers tomorrow when multimedia and telephony are added to existing backbones.But no one now believes that ATM will irresistibly wipe away Ethernet, or its derivatives.
As Palmer says: "A year ago, a lot of people said ATM was the answer to everything, but fast Ethernet will be around for a very long time."
Case Studies: ATM at Oracle and BAE
As probably the UK's leading ATM evangelist Var, K-Net has experience of installing more than 200 networks based on the technology. Two of its most recent coups involve British Aerospace and Oracle.
The Bae contract is unusual because they were installing a network from scratch, rather than upgrading an existing infrastructure. What K-Net could do was provide an ATM backbone to serve 1600 users.
ATM is the backbone for British Aerospace, because they demolished two of their exisitng buildings and built two more to replace them as part of their Fit for The Future project. They had decided that their management style was wrong and they needed to be project-focused to be able to build and break teams quickly, so ATM using Lan emulation was right for them," explained Andrew Rowney, ATM product manager at K-Net.
The installation shows that ATM doesn't only have to be sold on bandwidth: reliability is now good, and its data sharing capabilities are suited to supporting the modern organisation. "Since the network was installed in October last year, they have has a total outage of 10 minutes," Rowney added, "They are not interested in ATM, they are interested in the ability to reorganise their business."
Meanwhile back in the computer business, few companies have the need for bandwidth that Oracle has. And few have as many offices either. "Oracle has four offices in Bracknell, plus two in Reading and they wanted bandwidth at least equivalent to what they had internally when it came to linking the sites - and internally, they are FDDI users. What they now have is a private WAN with a system that can handle all their existing voice and data traffic," Rowney explained.
The campus Wan at Oracle now supports more than 2,000 staff, and they are treated to a galaxy of applications that show what ATM can be used for: a VPN, the hub for the Oracle Channel satellite TV service, plus a soon-to-come videoconferencing and distance learning application too.
If many users outside the computer business are ignoring networked multimedia, at least one inside the business is walking the talk.
If ATM is to deliver its full range of benefits, it will need to give users the opportunity to access voice at the desktop as well as just data.
But all the talk of convergence at the moment is just that - talk.
Garry Smith, convergence activist at CTI specialist Mitel - or product manager for ATM, if you prefer, is looking at convergence as tomorrow's technology rather than today's. "We don't deliver ATM to the desktop yet.
We look at just the backbone, but in conjunction with Intel and the USB initiative, we have developed telephone systems that work with that," he explained.
Smith was getting a lot of interest for his backbone products at the recent Networks Show in Birmingham. "Today we can converge the backbone - the PBX we manufacture isn't a big box in the basement - it is a hub, like the data hub. The current level of technology allows little more than the passing of the most basic information between users, but our idea is full multimedia transport round the organisation," he added, hopefully this may be a pipe dream for a while yet - in the endless series of ATM standards agreements, voice telephony has been at the back of the queue.
"The missing piece for us is a ratified standard for voice telephony over ATM," Smith said. Nevertheless, he considers the image of ATM as a series of proprietary technologies is damaging, and is working to convince potential customers that ATM will not be a cul-de-sac. "We have been trying to educate our users that most of the standards have been agreed. I would say that we will one day live in an 'ATM world', and adjuncts to ethernet - like switched ethernet - are ultimately not developed like ATM was, which from day one was designed to handle asynchronous traffic,' he added.
Vendor's announcements include AI-powered Microsoft Office, a move away from password verification and an alliance with Adobe and SAP
Vendor claims hackers are hijacking machines to mine for cryptocurrency
Nearly half of SMBs are planning to invest in digital workflows to reduce their paper-based processes by 2025, according to Quocirca
The charter has pulled together the biggest names in tech in an unprecedented attempt to address the tech industry's lack of diversity. Tom Wright asks how it plans to do it