Network management Toolkit

Bob 'Ethernet' Metcalfe looks to the future

Tags:
Bob Metcalfe,
Ethernet,
ZigBee,
Netevents

Graeme Wearden ZDNet.co.uk

Published: 11 Oct 2004 15:45 BST

Bob Metcalfe changed the IT world back in the early 1970s when he invented Ethernet though his work at Xerox's Palo Alto Research Centre. Thirty years on, his invention is ubiquitous in the local area network and is making big strides into the metro space too. Metcalfe went on to found 3Com, and is now a general partner at venture capital firm Polaris Ventures.

Last Saturday Metcalfe delivered a speech at the NetEvents conference in Barcelona, giving a visionary's take on the future of networking and the wider IT space. Metcalfe predicted that ZigBee will have a very bright future, warned about the problems of cramming so much technology into a mobile handset, and joked that whichever networking technology manages to usurp Ethernet will end up being called Ethernet.

ZDNet UK was there to get Metcalfe's take on the future of networking and technology.

Q: Which other technologies can have the success of Ethernet in the future?
A: You mean that won't end up being called Ethernet?

Will ZigBee be called Ethernet?
If I have anything to say about it...

We're pretty excited by ZigBee, and we [Polaris Ventures] have invested in several ZigBee companies, so we're now corrupted and you can't believe anything we say.

Frequencies are becoming available that have not been used much before, at between 50GHz and 100GHz, and they can be accessed by new chips that are coming. There's very high bandwidth up there. One issue with such frequencies is that they tend to be more dependent on line of sight, but that can be solved by software.

Mesh is another exciting technology. I mean, the Internet is a mesh, but all wireless systems will be meshes soon.

Can enterprises keep up with the increasing speed of networks? Do we need ever faster networks?
When 10Mbps Ethernet came along they couldn't keep up with it, and when 100Mbps Ethernet came along they couldn’t keep up with that either.

But in a system where you have clients and software and other components, it's a bad strategy to argue that "since everything else is the bottleneck then I won't improve what I do". The problem with that is that no-one improves anything very much, because it's never their turn.

If your attitude is "I'm not the bottleneck but I'm moving to the next-generation of my product anyway", then you'll have people complaining that "you don't need to upgrade".

The right answer is to keep innovating even if it is overkill at the time. 10Mbps Ethernet was overkill in the 1980s, but soon it was necessary.

What lessons can we learn from Ethernet being 'good enough' for networking rather than heavily deterministic as many of its competitors were?
This used to be a big debate when it was Ethernet versus Token Ring. Determinism was the issue. Ethernet no longer has collisions, but it never was a real problem because the whole issue was overplayed, often by salesmen making the best arguments for their products.

Being deterministic was not really an issue. With Token Ring, if you lost a token then you were very non-deterministic.

One of Ethernet's strengths was that because it came out of the Arpanet initiative it understood its place in the protocol hierarchy. So it did what it had to in that place, in levels one and two, and it was good enough to cope in the hierarchy.

Token Ring, though, competed too much with the protocols working at other levels.

As well as token ring being too expensive?
Twice as expensive [as Ethernet].

In your speech (at the NetEvents conference) you mentioned the problem of cramming multiple radios into one mobile phone. Is a software radio the answer?
Not yet, but it's another area that's coming. Software radios are one way of having many radios in one bit of silicon. There's no danger of it happening suddenly.

There's a company in the US called Vanu who are already doing software radios. It's run by a man called Vanu Bose, but his father had already used the name Bose for his own company.

They're beginning to have base stations based on software radios, but it'll be a long time before it is practical.