[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Wireless transfer throughput?

I wonder if anyone can explain to me why the achieved throughput for an rsync
transfer for files across a wireless connection should be different than I
expect as I try to explain below?

The situation is that I have a local LAN over which I can run a transfer
from two different laptops to a wired only desktop machine on which I keep
backup files.

The two laptops have different wired and wireless NICs.  On the wired
connection on a 10/100 system I get file transfers down the wire to the
desktop at about 11.7MB/s as measured on the gnome netspeed applet, which is
about what I would expect since this is around 95mbps and I could easily
imagine a 5% loss for TCP overheads. This is the same for both laptops.

On the first laptop if I do file transfers via wireless through its iwl3945
wireless card then I get an achieved throughput through my AP using WPA2
encryption at around 0.75 MB/s which corresponds to about 6mbps - and this
is on a link which the laptop system tells me is connected at 54mbps. The
machine is a few feet from the AP and shows a 100% signal.

On the second laptop which runs an iwl4965 wireless card I get around
1.5MB/s, corresponding to about 12mbps for the same conditions.

All machines are running F10, but these speeds were the same for F8 before I
did clean F10 installs on them all, so the specific version of Fedora seems
to be irrelevant here.

The way I look at what speeds I would expect is that I would imagine a
roughly (estimated) 50% loss due to the WPA2 encryption, so I would hope to
see around 27mbps, which corresponds to just over 3 MB/s. So I am seeing a
50% loss roughly on the machine running iwl4965 and another factor of two or
so on the machine running iwl3945.

Is this a problem with drivers or have I done my maths wrong? The machines
are running the stock kernel and drivers and currently am running
NetworkManager, though previously I was using wpa_supplicant directly.
Either way these details seemed not to change the numbers.

Any thoughts on this? I would be interested in being educated about this.
View this message in context: http://www.nabble.com/Wireless-transfer-throughput--tp21134634p21134634.html
Sent from the Fedora List mailing list archive at Nabble.com.

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]