The other day I was at a meeting about the plan to roll out an IT network at the Druk White Lotus School. It’s a challenging environment – 40 below overnight in the winter, nearly a meter of snow is typical, the ground is mostly dirt/sand/dust and rock, the site is about 3500m above sea level and the UV rays are enough to bleach anything made out of plastic in a couple of months. Electricity is hard to come by, but fortunately the site has a reasonable supply thanks to the PV system installed in 2008 and a trusty diesel back-up generator.
So the challenge is this – how to distribute network access across a 1km long by 500m wide campus with distributed buildings?
My preferred option was a co-ax 10Base2 network, based on an erroneous recollection that co-ax could go a ‘long way’ i.e. about 500 meters. Wikipedia soon put me right – the actual spec limit is between 180 and 200 meters. It’s also prone to total collapse in the event that the cable fails in a single point, or someone steals a terminator from the end! So probably it fails the fault tolerance test by some margin as well.
Another strong contender would be to place Wi-Fi repeaters on the top of each building – they could also be solar powered, so that the whole system would be ‘independent’ of any fixed infrastructure. The theory would be to connect just one to a local broadband connection and then share this across the site. In theory it’s a neat solution, however the availability of an off the shelf solution is limited. I’ve found a couple of attempts to delivery such a system – the closest is Green Wi-Fi but they don’t seem to have anything more up-to-date than a mod to a Linksys WRT54 router.
We thought about fibre, but the termination equipment is currently too expensive to be in the right ball-park.
After some more thinking, I’ve convinced myself that Powerline might be the answer, especially now that adapters are available for 2.5Mbps throughput) on the campus then it would be a significant time/money/effort/maintenance saving in comparison to laying new cable infrastructure.
Subject to the outcome of the powerline experiment, it seems to me that the longer term future is in the development of an ‘internet in a crate’ rooftop Wi-Fi solution – just like the one developed by Green Wifi, but implemented in a way that isn’t subject to manufacturer product cycles. Home routers are subject to repeated product cycling, with the hardware being squeezed smaller and smaller each time – as evidenced by the Linksys WRT54 no longer being capable of supporting OpenWRT. Whilst there are still plenty of OpenWRT-able routers out there, it would be nice to have a solution that’s platform based and sufficiently adaptable to survive product cycles.
A good ‘architecture’ for such a system would be a Linux based SBC (like the Beagleboard) with a USB hub and some wireless USB adapters. The adapters would have to be carefully chosen to provide Linux drivers and an MCX/equivalent connector for a bigger antenna, but in principle there will always be something close enough to this on the market. The system would be composed of:
- An SBC (such as, but not necessarily the Beagleboard)
- Linux (with some routing running on top)
- Solar Panel (ideally something small and cheap)
- Battery & Charge controller (if overnight operation is required, temperature considerations for the Druk White Lotus School might make this challenging)
- USB Hub
- Wireless USB adapters (minimum of 2 No.)
- Directional Wi-Fi antennae (minimum of 2 No.)
- Robust box to put everything in.
So in summary I bet you could build a prototype for about £200. The emphasis would be on using modular, off the shelf hardware that can be easily programmed and customised – but most importantly that each component can be potentially swapped for alternatives without the need to radically change the whole design. My main sticking point at present is discovering an SBC that’s sufficiently cheap and basic but capable of running Linux and supporting USB host functionality.
Linux and routing software isn’t my speciality, but I’ve discovered XORP which looks like a rather nice potential solution to the complexity of routing. It’s based on FreeBSD, but they say it compiles on Linux, so I guess that should work.
Next? I’m writing a one pager summary of what I’d like. If anyone wants to take this on as a project then let me know! The idea’s been around for a while, but nobody that I’ve looked at seems to have ‘cracked it’ yet. Hopefully I can persuade a university student somewhere to take it on as a final year project!
Post Script –
The Netgear WNR3500L looks like a good potential alternative to using the beagleboard as an SBC solution. Will need to see if it’s theoretically possible to mount a second Wi-Fi adapter using it’s host USB port.