The more I see of large-scale WiFi, the more it seems to me that the Mesh approach is the only way of doing it.

And yet, there seem to be people determined to talk it out of existence... I'd love to see it tried on some of these projects!

Edited by author 01-18-2006 02:34 PM
Agreed mesh is definitely the way to go but the problem with a "classic" mesh is that its implementation results in huge performance problems across multiple nodes so there's no way of getting anything like a decent quality of service on it. I know, beause I've tried it. We tried an implementation of voice calls on the open Linux meshAP system, which incidentally a lot of other companies have contributed to - not just one. Well the performance losses we measured were up to 90% degradation across the furthest apart bits of the mesh - we compared this to the - much more quietly developed Cambridge Matrix system, where the maximum losses we got across the network were 15% which is much better for voice calls...

George, two questions! - have you used the Locustworld dual-radio mesh? and what about Strix Systems?

Guy your link is dead...

I would have to concur wth George though I can't see traditional Mesh scaling effectively for Metropolitan area coverage.

We are in the middle of tender process for a Citywide deliverable so I can't comment too much but will post again once I'm able to and explain our approach.

Guy wrote "have you used the Locustworld dual-radio mesh?"

until such time as Locustworld breaks out of its coma and delivers working 802.11g (and 11a) functionality on a competitive range of wireless cards I would question whether it's a credible player in today's market. Adding another radio, or using structured wireless backhaul, can help but the fundamental limit of about 6 Mbits/s of traffic on an 802.11b radio is a killer in today's marketplace.

That 6M has to do inter-node mesh traffic, end user uploads and end user downloads, all at the same time. How can an 802.11b mesh operator compete with 2M ADSL let along the "up to 8M" and "up to 24M" services becoming available. How does an 802.11b operator utilise one of these fast DSL services or bonded DSL as a backhaul from the mesh if the meshbox node it connects to has less capacity than the feed ?

Sitting back and waiting for others to come up with 802.11g card drivers while using an old Linux kernel version (2.4.20) doesn't impress me. I have an 802.11g Prism54 card running in master (Access Point mode) in a Linux box, off the shelf SuSE 9.3, what does it need to do to be capable of running a mesh system ?

The Locustworld mesh concept is excellent and the WIANA management tools help make it into a workable scalable system, but it is a skyscraper built on a sand dune if it continues to be limited to 802.11b.

Community networks are having to add 802.11g overlays to their systems using WRT54Gs and the like, to try an offer a competitive service and offer their customers faster speeds.

Have Locustworld lost the plot ?

Phil