"People don't buy network switches any more," says Mark Canepa, the CEO of Extreme Networks.
Fortunately for him, as the boss of a company that specialises in high performance network switches, what he means is they don't buy them for their own sake.
"They have business problems to solve, and the switch is part of that," he explains. "For example, they're doing BPR, with new applications, and the current network is not able to absorb those, there's security problems...
"When I talk to a CIO, we rarely talk about networks, we talk about applications and the ability of the network to host those. The network is an integral part of the application now, and security is part of that – but people immediately associate the network with security problems too, such as denial-of-service attacks or viruses."
In Canepa's - and Extreme's - view, today's switches are not merely about connectivity, they are about applications and services embedded in the network and available at each port.
The network is the services
"How we think of it is every port is a proxy for services or features," he says. "It's a Gigabit port with 'this' list of features behind it. The list of features will probably grow forever, but today the port can for example take your username and password, and check it against a list, for 802.1X level security.
"The terms we use are insight and control. Insight means every port is intelligent enough to make judgements on a packet – is it clean, do we route it to a scanner, or do we prioritise it? Control is the ability to use the knowledge you gained from that to pass it on to a device with more insight."
He calls this 'multidimensional Ethernet' but says that in the current market, the real activity and opportunity is on the carrier side, not the enterprise side.
"Enterprises don't need multi-dimensional Ethernet nearly as much as cable companies do," he explains. "For example, triple-play [phone, TV and broadband] at $100 a month - the moment you have that, song downloads can't have the same priority as your IP phone or TV stream.
DSL vs Ethernet
"We have finally invented a technology cheap enough to get Ethernet to the home without needing DSL. For example, there's a Scandinavian company which has built a flat Layer 2 network that it leases to TV content distributors - it needed 10/100Mbit/s at the right price per port. DSL is not actually that cheap and we can now build IP for the same price. What we really need though is fibre-to-the-home."
That's in large part because of the bandwidth involved in delivering TV over IP. Canepa gives the example of Italian telco Wind, which uses DSL and therefore has to switch channels at base, because one channel chews up 8Mbit/s. He says Extreme made that possible by getting the latency right down.
"But if you want more than one channel at a time, you need another DSL," he adds. "So now they're looking to fibre – 100Mbit/s means 10 channels, maybe more with compression."
For telcos such as Wind, the big advantage of DSL is its subscriber management capabilities. Canepa admits that those don't yet exist for IP, but points out that the switch produces XML reports, "so you could write an application to look into an SQL database and program the switch to turn the flows on and off.
"Now though you have security problems – and the least of them is people hacking the system to get free TV. What if they take the whole network down? That's why we build things to stop laptops getting onto the IP telephony VLAN, for example."
Changing the V in VOIP
He adds, "Our VOIP strategy is really simple – we build converged networks that are voice-grade. We have a number of partners, for example Avaya is furthest along and is strong in the enterprise space, but we also have programmes with Mitel, ShoreTel and Siemens. We support seven or eight phones and we helped invent LLDP.
"The question is what's going to be the next thing for the network? It was voice – that's latency, not bandwidth. Video is interesting – HDTV is a bit of a challenge, while latency and isochrony are issues - people will be less tolerant of poor quality TV.
"I think VOIP will become video-over-IP, with telepresence – maybe in SuperHD on a 200-inch screen in your house.
"People are watching video on phones, doing telereporting, with kids chasing the police and monitoring them. People are illicitly streaming video from boxing matches. The technology is coming for button-sized cameras with their own IP address, for $1 each. You can already buy IP phones for $15 each if you're willing to buy 5000.
"I would be surprised if in a few years you can't get a wireless webcam that doesn't need a battery. They you have all the security and access problems..."