Have we gone MAN overboard?

Vendors struggling to make the case for metropolitan ethernet networks...
Written by Jon Bernstein, Contributor

Vendors struggling to make the case for metropolitan ethernet networks...

"Carrier backbones are at 10 per cent utilisation. Who wants to install a bunch of terabit switches when only 10 per cent of the network is being used?" Listen to Gartner's chief analyst Ian Keene and you'll wonder why so many big name vendors - from HP to Mitel, from Intel to Nortel - are clambering to be part of the gigabit (let alone terabit) gig. And why they are trying to take their resolutely campus-based technology out into the big, bad world. For a highly complex technology, Gigabit Ethernet across the MAN (metropolitan area network) poses two very straightforward questions: - Why do we want it?
- Why do we want it more than the technology that already exists? There are, of course, more than one 'we' in these questions: the 'we' the user, and the 'we' the supplier. Both are currently stuck on question one. Granted, no networking protocol can deliver anything approaching 100 per cent utilisation, but 10 per cent? As Keene points out, the fibre laid in the ground is waiting for the application demand to spark it into life. Ask a networking vendor what that application is going to be and they'll give you myriad of answers from video-conferencing and streaming to apps on tap and distributed storage. Others suggest users just want to do the same things - resource and file sharing - only more efficiently and across increasingly wide boundaries. Some technologies - like videoconferencing - have wrongly been cited as killer applications for so many years that it's difficult to believe their time is now. Others like apps on taps - and the application service providers (ASPs) that deliver them - are in danger of going the same way. Boring as it may seem, storage may be the one that makes sense of the metropolitan area network. Not a killer app with quite the cache of email or text messaging but if predictions that before the end of this decade we will create as much data as the entire human race did in its existence hitherto, then data storage could be the challenge that high-speed networks help to meet. Having accepted that there may be an application out there to fit the bill, what makes us think Gigabit Ethernet is the technology to fit that bill? Critics, such as Eolring's director of marketing Victor O'Neil point out that Ethernet wasn't built for long distances, and it shows. "Ethernet has a number of difficulties," he says. "It does not have clear and precise bandwidth management capabilities. It does not have clear and precise guarantees in terms of levels or quality of service. It does not have very precise response times in terms of network resilience, network uptime and availability. "These are the all issues that a metropolitan area operator will have to address if they are going to generate revenue out of customers." Nothing here, it would seem, to challenge the incumbent technologies, most notably Sonet/SDH (synchronous digital hierarchy), a transmission standard used by most of the world's carriers today. The Ethernet camp already acknowledges some of these drawbacks and is working on two standards it hopes will assuage user fears. Namely, Resilient Packet Ring (RPR), which should solve the first problem and ITU x.86, which addresses the second. Brice Clark, director of strategic planning at HP, says O'Neil's concerns are overstated. "I do not think it is relevant to ask whether or not Ethernet is the best technology. It is the most installed technology from a data perspective which is part of what is driving this out of its origins. It holds the promise of a much more unified fabric." Gartner's Keene is even more pragmatic in his support for Gigabit Ethernet. "Metropolitan networks, in the new optical age, are going to be about throwing cheaper bandwidth at the problem, not very sophisticated bandwidth, but very cheap." The cost to the carrier of installing a Gigabit Ethernet network is 20 per cent of a Sonet/SDH network, according to Keene. If he has his doubts about the overall use of carrier fibre, he does at least believe there will be some demand. Gartner predicts, with an 80 per cent probability, one in five medium and large businesses around the world will have at least one metropolitan area connection in their enterprise by 2005. And while over half of all carrier equipment will remain Sonet/SDH based by that stage, it still leaves plenty of room - and hope - for Gigabit Ethernet suppliers. That's assuming users can answer the application question.
Editorial standards