Last week, Tropos' launched a new version of its MetroMesh software, designed to make Wi-Fi mesh networks more suitable for service providers - as envisioned by the municipal network movement.

But, there is considerable discussion about how far Wi-Fi mesh can gotowards building a reliable infrastructure, when it relies on shared, bandwidth in the unlicensed bands. Some have argued that carriers won't really support it.

Tropos disagrees. We spoke to Bert Williams, marketing vice president at Tropos, to find out why, and what Tropos' long term plans are.

Integration with carriers
"We're now moving to the third phase of Wi-Fi mesh - to the point where large service providers carriers can implement these networks successfully, with a nationwide footprint," said Williams. "We have tighter integration with carrier bandwidth management and authentication systems."

There are three main improvements in the new version, all aimed at making Wi-Fi mesh carrier-class, he says:

  • Increased capacity
  • Rate management, and
  • Links to authentication systems.

"Our previous version gave 10 to 15 Mbit/s concurrent bandwidth per square mile," said Williams. "Now we offer 20 Mbit/s." The increase is provided by a software upgrade, not a hardware replacement, he says, so it is available to anyone paying a maintenance contract for existing Tropos sysatems.

The increase comes from automatic channel allocation, in which the network will develop its own plan for which Wi-Fi channels are used at which nodes, and this can be optimised occasionally. The network channels are fine tuned over time, by dynamic channel assignment - "an enhancement to our predictive wireless routing protocol," says Williams.

Rate management is vital if service providers are going to use Tropos, because they have to be able to limit some users, in order to preserve bandwidth for others. Tropos systems could already share the uplink to the Internet, he says, but the bandwidth manager was helpless between users who are neighbours on the wireless network."

The new version can limit connections between devices on the mesh. "If users are doing peer-to-peer traffic amongst themselves, we can prevent them from hogging the network," says Williams. "This gives operators the option of charging more for a better service. Before could only do bw mgt at interface between wireless and wireline. That was useful for Internet traffic, but led to possibility . Push bw policy manager out to the edge of the network. If trying to get more bw than their SLA supports, would be stymied.

Finally, the new version supports the authentication systems used by service providers: "We integrate seamless mobility with authentication systems," says Williams. "We use tunnelling and BGP routing to make the core routers mobility aware."

"You can have a network in Podunk and an authentication system in Peoria," he says. "Users can go anywhere in Podunk and the authentication system will keep them authenticated and know where they are." They can move round the network, handing off between access points, without losing their session, he says: "That's key to allowing operators to provide large scale mesh networks."

How about the spectrum?
But, we asked, what about the fact that the Wi-Fi spectrum is unlicensed? All the service guarantees in the world won't help you when someone else sets up a WLAN in your space - and if it's unlicensed, there's not a lot you can do then, is there?

"Those concerns are way overblown," said Williams. "We haven't seen them borne out in the field." He puts his trust in Tropos' channel planning, which is "very sensitive to and reactive to interference."

"Licensed spectrum is no panacea," he went on. "There is still interference in licensed spectrum. There are places you go and your call gets dropped, either because there is no coverage or because there is interference. It's important to realise that Wi-Fi was designed to allow different networks to share the same airwaves."

Wi-Fi doesn't suffer from a "tragedy of the commons" problem, he says, where a resource gets used up because no-one owns it, and therefore no one has any incentive to conserve it. "That's just not the way WiFi works," he said. "As more people use it, the bandwidth per user goes down, but the total bandwidth does not. The protocol was designed to share the airwave even if there are many diff networks. Wi-Fi can be shared amongst providers efficiently."

He's being a bit disingenuous, of course - he's just admitted that Wi-Fi bandwidth is a finite resource, and when more than one person share it, everyone gets less. Whether or not it's a "tragedy of the commons", it's not the same problem as in the cellular spectrum, where interference is limited by the number of licence holders.

His example of how good Wi-Fi is at sharing misses the point: "In our office, we have around 150 MetroMesh routers running in a 20,000 sqare foot buiding, and it works just fine," he says. Well, of course it does - it's all part of the same mesh, run by the same provider.

Future protocols
In fact Tropos has a plan if it turns out providers need licensed spectrum - WiMax. But it's not going to come overnight.

"About a year and half ago, we set out a three-phase WiMx strategy," he told us. "Phase one uses point-to-multipoint for backhaul, while phase two puts the backhaul in the mesh, and phase 3 puts WiMax in the client. We were pooh-poohed for having pessimistic time-frames, but I think they may have been optimistic."

"People are stating to understand that the notion that if you stick a WiMax tower in the centre of town, everyone within 30 miles will automatically have broadband, is pie in the sky," he said. "When it eventually reaches the client WiMax will be delivered through mesh."

The density required will drive operators to mesh, he believes: "In terms of Mbit/s per square mile, you really need to have a dense cellular architecture. Once you've realise you have to have a lot of nodes, it automatically pushes you to mesh. You can't pull wires for 10 to 30 base stations per square mile."

Williams believes that this year, WiMax vendors will bring out product suitable for backhaul in an outdoor Wi-Fi mesh.

When it comes to Wi-Fi flavours, he's quite excited about 802.11n, which he expects to hit mesh networking in about a year. But 802.11a leaves him cold, even though some have suggested that the uncluttered 5GHz spectrum would be very good for mesh backhaul.

"We have a somewhat different view," he said. "We use a lot of 5.8 GHz point-to-point links today, to offer connectivity from the mesh to the point of presence, but we have chosen thus far not to build it in as backhaul within the mesh. 5.8 GHz does not propagate well out of doors." Ironically, 2.4GHz would be better for backhaul, but 802.11b/g is built into all the clients.