To: djane who wrote (43857 ) 4/12/1998 2:36:00 AM From: djane Read Replies (4) | Respond to of 61433
Is ATM doomed? High-bandwidth technologies duel for network dominance internettelephony.com WAYNE CARTER, 4/6/98 As dense wavelength division multiplexing multiplies the available channels on a single fiber strand and terabit routers boost the capacity of those channels, talk has begun to surface that asynchronous transfer mode's role in the network may be short-lived. The argument is that the enormous capacity provided by DWDM--or ultra-dense WDM, as some newer systems are being called--and terabit routing eliminates the need for a traffic-grooming technology such as ATM. That would be a boon for Internet protocol (IP) networks and the vendors that supply IP-switching equipment. IP traffic wouldn't need to be mapped to ATM to provide the quality of service (QOS) necessary for voice or fax over IP and video transmission. Milo Medin, networks vice president for @Home, has predicted that ATM switching equipment between routers or switches and WDM equipment will be made obsolete by multiple OC-48 interfaces (Telephony, Internet Edge, March 16, page 37). David Passmore, president of NetReference Inc., recently echoed that sentiment. "There's not much of a role left for ATM," he said. "Once you have terabit routers that are carrier-class products, you can link them together using Sonet or WDM. ATM just complicates things." Not everyone, however, is convinced that ATM is on the way out. Companies that have staked considerable investments in ATM equipment and networks certainly have a vested interest in keeping the technology afloat, but even some of those companies see a changing role for ATM as the network evolves. Xylan Corp., which established itself in the LAN environment before moving its ATM OmniSwitch to the carrier market, believes that it is well-positioned to capitalize on ATM's evolution. ATM was developed by carriers to get the most out of their fiber networks, especially long-haul routes, said Richard Henkus, Xylan's carrier marketing director. "It was thought that ATM would be the ultimate core network," Henkus said. "Everything would be multiplexed to ATM. It was efficient use of scarce resources." Henkus agreed that the advent of terabit optical devices and DWDM makes bandwidth far more available than when ATM was developed. But a need still exists for ATM in the access layer, where local carriers face the challenge of continuing to serve traditional voice customers while adjusting for the increasing flood of data traffic. "We see a modification of the role for ATM, but ATM is still the best statistical multiplexer," he said. New telecom equipment vendors are facing the same questions. Avici, Juniper Networks, NetCore and Nexabit will introduce switch routers this year aimed at delivering QOS capabilities to data-centric service offerings and carriers. Nexabit hasn't released its architecture, but its plans are to make IP the primary transport mechanism--even mapping ATM cells to IP for transport. On the other hand, NetCore is sticking to the more traditional strategy of relying on ATM as the mule that hauls the traffic. ATM is the only way to truly ensure service quality over data networks, said John Shaw, NetCore's marketing vice president. The multiprotocol label switching standard under development may bring IP QOS up to the level of ATM, but "until a standard approach to QOS is implemented to IP, it becomes a tenuous step to provide QOS," he said. IP QOS systems are not close to providing the same performance as ATM schemes, said John Coons, director and principal analyst for Internet infrastructure at Dataquest. "IP QOS systems all sound the same," he said. "If you mark some traffic as premium traffic, it may be given preferential treatment, but they can't guarantee [service levels for] delay or packet loss." And IP QOS is a tough sell to end-users as well, he said. Shaw said ATM will hold a place in the network hierarchy even as other technologies evolve.