David Young wrote:
> Do you expect much more than 24 Mb/s?
I certainly do, see these results (this is in turbo mode):
802.11Ta (ch 152):
Bit Rate=54 Mb/s : 23.5Mb/s
Bit Rate=36 Mb/s : 23.5Mb/s
Bit Rate=24 Mb/s : 23.9Mb/s
Bit Rate=12 Mb/s : 17Mb/s
Bit Rate=6 Mb/s : 9Mb/s
Note how bandwidth is roughly 1.5 times the link bitrate, so I ought to
have been able to do about 81 Mb/s at 54Mb/s.
> According to my back-of-an-envelope
> computation, you cannot expect much more than 28 Mb/s with 1500-byte
> packets and a 54 Mb/s 802.11g link, which is about what I got under
> "ideal" conditions.
Hmm, how does that work?
> How do you know that the CPU is the bottleneck? Sometimes CPU usage
> measurements are not reliable.
CPU idle as reported by top is 0% as soon as 24Mb/s is hit, in both
turbo and non-turbo modes.
> What iperf parameters do you use?
None, really.
> Are you measuring TCP or UDP performance?
That would be TCP.
> If TCP, are you sure that your window size is large enough?
> Do you have enough buffers
I dunno, how do I check that?
> Is the 4801 the "source" or "sink" for the data stream?
No, because iperf takes a bit of CPU.
> If not, does
> the 4801 route or bridge from the wire to the wireless?
I use plain routing, which is very close to bridging wrt performance.
> Have you tried fixing the wireless transmission rate to the very highest?
That's what I was doing to get the performance numbers at different
bitrates.
> Can you isolate the devices under test from interference?
I have, there are no other 802.11a networks in range.
--
Regards Flemming Frandsen - YAPH - http://dion.swamp.dk
_______________________________________________
Soekris-tech mailing list
[email protected]
http://lists.soekris.com/mailman/listinfo/soekris-tech
