Hello Helmut,
thanks for pointing that out!
As far as I understand minstrel, it is using sent packets and ack statistics
to find out the percentage of successfulness. Therefore, minstrel should
probably not consider NOACK frames for its statistics.
If there are only AC_VO frames, there would not really be any data for
minstrel to build its statistics, so one could either rely on the
control traffic which may go over another queue or use fixed rates.
For example, SIP uses RTP for the voice data and SIP packets for
session control ...
best regards,
Simon
[Helmut, sorry for the resend, I accidently replied personally to only you first]
On Wed, Nov 16, 2011 at 02:58:18PM +0100, Helmut Schaa wrote:
> On Wed, Nov 16, 2011 at 12:21 AM, Simon Wunderlich
> <simon.wunderlich@xxxxxxxxxxxxxxxxxxxx> wrote:
> > This patch adds support for NoAck per WMM Queue. The Unicast QoS
> > Header is adapted accordingly for each outgoing frame.
> > The support is turned on and off through nl80211 by extending
> > the WMM TX Queue Parameters, but can be triggered separately.
>
> Just a thought: Assuming you've configured AC_VO to use NoAck and
> you've only got VO traffic going on, is minstrel still able to collect enough
> statistics to do proper rate selection?
>
> Thanks,
> Helmut
>
Attachment:
signature.asc
Description: Digital signature
