I have implemented the following mechanism to measure the transmission
time in madwifi.
First of all I set the ATH_TXQ_INTR_PERIOD to 1 to receive one report
for each packet sent.
Then I measure the transmission time in two different way based on the
status of the transmission queues.
If the transmission queues are non empty, I measure the transmission
time as the time between two consecutive ACK reception.
Otherwise, I evaluate the transmission time as the difference between
the ACK reception time and the time when the driver insert the packet
into the transmission queue.
Probably this isn't the optimal way to evaluate the transmission time,
however the transmission time
I obtain is accurate enough to infer the network status.
Ing. Emilio Ancillotti, PhD
Pervasive Computing & Networking Lab. (PerLab)
Institute for Informatics and Telematics (IIT)
National Research Council (CNR)
Via G. Moruzzi, 1 || voice: +39 050 315 2437
56124 Pisa, Italy || fax: +39 050 315 2113
|| mobile: +39 328 2963760
On Thu, Jun 25, 2009 at 10:03:21AM +0200, thus spake Emilio Ancillotti:
> Probably this isn't the optimal way to evaluate the transmission time,
> however the transmission time
> I obtain is accurate enough to infer the network status.
Very clever indeed. Nevertheless, doesn't setting ATH_TXQ_INTR_PERIOD to 1
incur too much performance penalty?
To err is human, to purr feline.