Up until very recently, the received wisdom of most networking industry luminaries was that the future of high-speed data communications belonged solely to Asynchronous Transfer Mode. Customers, however, have had other ideas. While ATM is winning some support in the wide area network market, it has spectacularly failed to penetrate the corporate backbone and seep down to the desktop. Proponents of ATM are increasingly struggling to convince skeptics that its costs will shrink sufficiently to make a switch from the workhorse local area network standard, Ethernet to ATM worthwhile, despite the technology’s advantages in terms of speed. Perhaps more importantly, users are also proving reluctant to adopt what is perceived as an untested environment. As ATM vendors have struggled to get their message across, the sizable Ethernet camp has fought back with a vengeance. First came fast Ethernet, running at 100Mbps. Now an even faster variant, capable of 1Gbps capacity, is being touted. Hopes are high for Gigabit Ethernet. The market research group Dataquest predicts that it will beat ATM to become the dominant communications backbone technology by the year 2000, creating a $2.9bn worldwide market in the process. In April, the IEEE announced that it had begun formal work on defining a Gigabit Ethernet standard, although that is unlikely to be finalized before the end of 1998. Meanwhile a Gigabit Ethernet Alliance was set up in May to spur commercial uptake of the technology in a dvance of the full standard, with UB Networks Inc, Sun Microsystems Inc, 3Com Corp, Compaq Computer Corp and Intel Corp among the 11 founders (there are now over 50 members). The group’s goals include resolving the technical problems that may hamper commercial acceptance and, with its own ‘draft’ standard due in January, it predicts that ‘pre-standard’ commercial Gigabit Ethernet products will appear as early as next year. Two companies that were also founding members of the Gigabit Ethernet Alliance – Cisco Systems Inc and a small research and development company called Granite Systems Inc – now only need one seat at meetings. Growth-hungry Cisco recently made its fifth purchase of 1996, snapping up Granite Systems in a $220m share exchange.
By Gary Flood
Cisco is happy to admit that the reason it was prepared to spend that much money for a 50-strong Palo Alto, California company that is years away from product delivery was to grab one of its founders, Andreas Bechtolsheim. Bechtolsheim’s career got off to something of a good start: he built the original Unix workstation while finishing his doctorate at Stanford University, and then left in 1982 to co-found Sun Microsystems Inc with Scott McNealy, Bill Joy and Vinod Khosla. Other milestones include doing much of the foundation work in building the RISC SparcStation. A genius with a short attention span, as Sun chief executive McNealy describes him, Bechtolsheim quit his role as vice-president of technology at Sun only last year, setting up Granite Systems – dedicated to ultra-high performance local network switching technology development – with $2m of his own money. Granite’s original mission statement was to simply offer high-performance network switches; the focus on Gigabit Et hernet came later, although there is speculation that fellow Granite founder David Cheriton’s work on VASSA Virtual Address Space Switching Architecture might be a factor. Bechtolsheim’s aim now, and one which Cisco clearly agrees with, is to develop Gigabit Ethernet in ASICs. The idea is that using this approach, Gigabit Ethernet can be deployed in a modular fashion, enabling Cisco, which has tended to concentrate on software rather than hard-wired implementations, to have greater flexibility in developing products. Other Gigabit Ethernet players, like Plaintree, are also working on ASICs, and the company previewed a switch claiming 16 Gigabit Ethernet ports at last month’s Networld+Interop. Given that Bechtolsheim is unlikely to have a ccepted Cisco’s offer purely for the money – which he does not need – the move must sound something of a death knell for ATM’s chances at the desktop level. The news that Bechtolsheim has committed to remaining at Cisco for at least two years, acting as the head of its workgroup business unit (which will be based around Granite Systems), suggests that Bechtolsheim has become convinced that, with Cisco’s backing, he will have the ability to dictate the future of communications technology to the same extent as his liaison with Sun helped set the mould of open computing. ATM developers may now have a truly formidable opponent on their hands.
