From: Brian Capouch (brianc_at_palaver.net)
Date: 2002-04-30 08:48:40 UTC
I am going to need to do some more testing here, but I have a
disconcerting report to give on a comparison between a Mikrotik 200mw
Prism2 card, running the latest version of HostAP, compared to a 30mw
Lucent White.
I'm sure there is a logical explanation, and I hope perhaps people can tell me where to look for my problem.
The test used a Lucent access point feeding a downtilt omni, which is the source of b/w for my usual network connection. I set up a laptop about 1500' from the omni. The Lucent card was using the wvlan_cs driver; the Prism2 used the HostAP driver, set in "Mode 2" which I think is STA mode.
The Lucent, using iwconfig, showed a signal strength of 30/92, got ping times to my test machine of approx 3.5 ms, ~500kbs throughput, and 0% packet loss.
I pulled that card, inserted and config'ed the Prism2, and got a 60/92 rating from iwconfig. I thought, "How marvelous." Not exactly. Throughput to the same place was ~220kbs, ping times were widely variable from 6.4 to 15.5 ms, and packet loss, in repeated testing was close to 50%.
I repeated the tests several times over, and the results were nearly identical.
I was freezing, by the way, and had to cut the test short. One thing I noticed while looking around in the /proc filespace was that using the HostAP driver there were thousands of octets of multicast showing having been received, and almost no unicast.
The bandwidth tester was iperf.
Can anyone tell me a) what might have been the problem, and b) what I can do to better nail down where the problem might lie.
The Mikrotik card had two extern antenna connectors, and even though I have always heard it matters which one is plugged in, my results were basically the same regardless.
Thanks. Sorry for all the words.
B.