Blanket-coverage wireless via desktop machine wireless cards in "Master" mode?

[dhicks]

Hello All,

 

We have a desktop machine in each classroom with spare PCI / PXI Express slots or USB ports. Shouldn't it be possible to simply install a wireless device capable of "master" mode in each desktop machine if we wanted wireless access in each classroom? It shouldn't be too much trouble to have a small client application on each machine that reported back to a central server so the channel numbers could be adjusted to reduce interference between adjacent wireless points. Am I missing some obvious limitation of such a system?

[dhicks]

We have a desktop machine in each classroom with spare PCI / PXI Express slots or USB ports.

 

Hmm - I've tried two different USB wireless devices, neither of which seem to support master mode. I've ordered an Asus PCE-N13 PCI Express card which certainly looks like it should work. I'm guessing USB wireless devices just don't have the space for additional functionality?

[SYNACK]

Reasons for eekness:

Bandwidth shareing (a class load of laptops and a pc on 1GB/s)

Security and config - All traffic passes through an easily compromisable device which must tie its interfaces together (fine on *nix but hell on Win)

Rubbish hardware: most wifi cards are very simple and so not will just spend the whole time jumping on the worstations interupts many also have low power radios less useful for AP type stuff

Always on (power usage)

Management - you would need to build your own system to manage it all, possible but tricky

Latency - All traffic queued up behind a non dedicated OS and its own needs.

 

If doing this on Windows you would need to write your own service that acted at a rather low level to share out IPs (ICS would be nasty due to horrific multi level NAT. (edit: perhaps a vlan subinterface bridged to wireless, still limited to one WLANID and unstable though).

 

In saying this with the right cards and the right behind the scenes system that could isolate wireless traffic and pipe it off to a seporate VLAN unmolested you could get a decent enough system but it would be complex and less robust than a dedicated system.

 

 

I am interested in what you come up with though, I had a similar idea several years ago but just setting up a couple to work reliably turned out to be such a mission that it became clear that without many considerations it was a huge time sink with limited results.

Edited June 17, 2011 by SYNACK

[dhicks]

Bandwidth shareing (a class load of laptops and a pc on 1GB/s)

 

True, but a dedicated access point is only going to have 1GB/s anyway.

 

Security and config - All traffic passes through an easily compromisable device which must tie its interfaces together (fine on *nix but hell on Win)

 

I'm going to sort out my home wireless access point first, which consists (hopefully) of that Asus PCE-N13 PCI Express card linked to a Debian virtual machine to do routing / filtering. Then I'll investigate how Windows does something similar.

 

Rubbish hardware: most wifi cards are very simple and so not will just spend the whole time jumping on the worstations interupts many also have low power radios less useful for AP type stuff

Always on (power usage)

 

I'll see how that Asus card is - you'd hope by now that all the useful wireless features would be available on an all-in-one chip that everyone uses. Low power could be seen as an advantage - less interference between stations, less wireless power for people to fret about. The access point turning off when the PC turns off should also be quite handy.

 

Management - you would need to build your own system to manage it all, possible but tricky

 

Indeed - it's a case of figuring out just how tricky. I'm rather thinking something to do with VLANs (or a VPN of some sort to a central server?) myself.

 

Latency - All traffic queued up behind a non dedicated OS and its own needs.

 

Hmm, good point.

[SYNACK]

True, but a dedicated access point is only going to have 1GB/s anyway.

Yea, you can never have too much bandwidth though and stacking a possibly high needs workstation on to aswell could limit it a bit more.

 

I'm going to sort out my home wireless access point first, which consists (hopefully) of that Asus PCE-N13 PCI Express card linked to a Debian virtual machine to do routing / filtering. Then I'll investigate how Windows does something similar.

 

I'll see how that Asus card is - you'd hope by now that all the useful wireless features would be available on an all-in-one chip that everyone uses. Low power could be seen as an advantage - less interference between stations, less wireless power for people to fret about. The access point turning off when the PC turns off should also be quite handy.

 

There are all types, some are better than others, many of the home grage WAPs actually use standardised chips, just look at something like the Tomato or OpenWRT project (which may give you some good ideas and code) which support these specific types of chips. They are still low end comodity chips though so have very little unloading which limits them or at least offloads hassle to the CPU. The output power and signal quality can actually be dictated a large amount by the external circutry and how much the signal is amped. You are right about the power levels but you need to have plenty of them. You may also want pigtails to raise the antennas from the back of the PC as metal is not overly permiable to WiFi.

 

Ideally you would want to use ABGN wireless cards with MIMO as spreading across the 2.4 and 5ghz bands along with MIMO will give you the triple the space and way more reliability in already wifi flooded areas.

 

 

Indeed - it's a case of figuring out just how tricky. I'm rather thinking something to do with VLANs (or a VPN of some sort to a central server?) myself.

 

 

VLANs would be better as then there is no encryption overhead and can be nicely and easily segmented, you could VPN it back in order to keep a flat network and bundle/unbundle the traffic each time but from a managment point of view its way more tricky any has huge overhead/delay compared the the realitivly menial task of tagging the frames for a VLAN and letting the switches handle it.

 

Ideally you want some way for the APs to talk back to a central server and detect each other and also detect and identify each others signals so they can self tune. A method to boot the clients between each one for forced host migration to help with load leveling would be another feature that most of the managed systems rely on.

[steveg]

All microsoft client operating systems are limited to 10 incoming connections, not sure if this would apply to ip traffic 'passing through' though, but if it does that severly limits what you can do with this.

 

I think if you factor in configuration and troubleshooting time, this is not going to be a cost effective solution when compared with a lower end managed wireless system.

 

Steve