On Thu, Oct 22, 2009 at 09:41:27AM -0400, thus spake Bob Copeland:
> Great, thanks.  In my environment I can maintain 54 mbps in an infra BSS,
> so I think it is dependent on the environment (or else there really is
> some problem specific to IBSS).  I do use IBSS frequently but only to a
> mobile phone that can do max 7 mbps so I haven't tried the higher rates
> there to be honest.

Yesterday I did several throughput tests with and without an 802.11g AP.

I had the two laptops with AR5212s, plus two D-Link DWL-G122 usb adapters
(driver rt73usb) for comparison purposes.

Test 1, ath5k in infrastructure mode:

                    | throughput | most common bitrates (DATA/ACK) in Mbps
  ------------------+------------+----------------------------------------
  laptop 1 uplink   |   9 Mbps   | 18/12
  laptop 1 downlink |   8 Mbps   | 18/?? (did not see any ACK in monitor)
  laptop 2 uplink   |  1.5 Mbps  | 12/11
  laptop 2 downlink |  7.5 Mbps  | 18/?? (did not see any ACK in monitor)

Test 2, madwifi in infrastructure mode:

                    | throughput | most common bitrates (DATA/ACK) in Mbps
  ------------------+------------+----------------------------------------
  laptop 1 uplink   |   0 Mbps   |  1/1  (almost all packets lost)
  laptop 1 downlink |  8.5 Mbps  | 18/?? (did not see any ACK in monitor)
  laptop 2 uplink   |   5 Mbps   | 12/12
  laptop 2 downlink |  7.8 Mbps  | 18/?? (did not see any ACK in monitor)

Test 3, rt73usb in infrastructure mode:

                    | throughput | most common bitrates (DATA/ACK) in Mbps
  ------------------+------------+----------------------------------------
  laptop 1 uplink   |  27 Mbps   | 54/24
  laptop 1 downlink |  31 Mbps   | 54/?? (did not see any ACK in monitor)
  laptop 2 uplink   |  27 Mbps   | 54/24
  laptop 2 downlink |  31 Mbps   | 54/?? (did not see any ACK in monitor)

Test 4, ath5k in IBSS mode:

                    | throughput | most common bitrates (DATA/ACK) in Mbps
  ------------------+------------+----------------------------------------
  laptop 1 -> 2     |  3.4 Mbps  | 18/__ (rate info from minstrel)
  laptop 2 -> 1     |  1.3 Mbps  | 12/__ (rate info from minstrel)

Test 5, madwifi in IBSS mode:

                    | throughput | most common bitrates (DATA/ACK) in Mbps
  ------------------+------------+----------------------------------------
  laptop 1 -> 2     |  4.8 Mbps  | 11/__ (rate info from minstrel)
  laptop 2 -> 1     |  650 Kbps  | ??/__ (lots of loss)

Test 6, rt73usb in IBSS mode:

                    | throughput | most common bitrates (DATA/ACK) in Mbps
  ------------------+------------+----------------------------------------
  laptop 1 -> 2     |  20 Mbps   | __/__ (rate info not available)
  laptop 2 -> 1     |  20 Mbps   | __/__ (rate info not available)


In tests 1-3, the usb adapter on the second laptop was put in monitor mode to
take a look at the bitrates of the frames.  In tests 4-6, I forgot that I
could have used the additional adapter as monitor on one of the laptops, so
the rate informations are taken from minstrel stats.

The rt73usb apparently doesn't support multi-rate retries, at least in the way
atheros allows it.  The TX interrupt is generated as soon as the frame is
pushed to the hardware, so there are no transmission stats available for the
rate control algorithm.  If you look at the minstrel stats, it appears that
it's transmitting at 54 Mbps with almost 100% success rate.

I also tried to do mixed IBSS tests (madwifi -> ath5k, ath5k -> rt73, etc) but
at some point I realized that channel 1 was quite busy and overall performance
was bad.  I then switched to channel 10, run tests 4-6 and run out of time for
mixed tests.

Generally, it's very difficult to get consistent results with my atheroses, be
it with madwifi or ath5k, sometimes the cards get stuck somehow, packets are
massively lost and the throughput falls consequently.

By the way, do you have any good experience with some atheros chipsets in
IBSS?  Would you recommend any specific version?  I'm thinking about buying a
pair of recent atheros-based USB adapters, on the conditions that they support
MRR the way mini-PCI atheros adapters do (I'm using the TX interrupt at the
end of real transmission to measure the service time of each frame).

Thanks for your help.

Ignacy

-- 
I have not lost my mind, it's backed up on disk somewhere.
_______________________________________________
ath5k-devel mailing list
ath5k-devel@lists.ath5k.org
https://lists.ath5k.org/mailman/listinfo/ath5k-devel

Reply via email to