Can anyone tell me a safe, *reliable* wireless configuration procedure
for this scenario?
We've got a scientific instrument running Linux 2.4.22-ac4 and an
Orinoco Silver PCMCIA card connected to an external antenna. The
instrument is going to be deployed in a remote location and visited once
every six months to download about 20 gigabytes of data. We need to
have the wireless working when we arrive at the scene, and it needs to
remain up until the download is complete.
Initially we had the card in Ad-Hoc mode, channel 1, and WEP key
enabled. iwconfig reported it as running at 5.5Mbps. It was stable for
a long time during testing, connecting only to my laptop with a Cisco
Aironet card similarly configured.
Last month we put the instrument outside the warehouse and began having
lots of trouble with the wireless. At times we'd see "eth1: Information
packet lost!" being blasted to the console. The computer in that state
is unusable: CPU pegged, unable to login, and unable to use the console
to issue commands to bring down the wireless or reboot the machine.
Then we'd start seeing messages about connections in the middle of the
night and access points coming in and out of range. The connection
between the laptop and instrument would die at random moments,
particularly if I saturated the link with a file transfer. If I took
off the WEP key the instrument would immediately see a connection but
the laptop wouldn't. If I *changed* the WEP key the instrument would
still see a connection from an unknown source. Once I got both
instrument and laptop to communicate, I could usually get a few days of
use out of it before things would go haywire again.
After much experimentation, I have implemented the following which seems
to work, but I'm not sure WHY it is more stable:
1) A wireless access point (Microsoft MN-700) has been provided with no
encryption whatsoever.
2) The instrument has been configured to use Managed mode,
essid="<access point network>". It detects the access point immediately
and seems to work.
3) Both machines have at the IP layer MTU=500 (instead of the default
1500) and TXQUEUELEN=10 (instead of the default 100).
I trust that the instrument is secure at the OS layer, so I don't need
WEP. I don't need the fastest possible speed either. When we go on a
maintenance operation, we just need to download the data reliably but we
can't guarantee physical access (hence the need for wireless). The plan
is to bring the access point with us when we visit the instrument.
My main question is, what is the *simplest*, most *reliable* way to get
wireless to work? Can I go any simpler than this?
Should I update the kernel on the instrument to 2.4.27? I am at
orinoco.c version 0.13d. Is 0.13e significantly more stable than 0.13d?
Any help is greatly appreciated.
Similar Threads
Linear Mode
