Re: [PATCH 7/7] mac80211: improve minstrels rate sorting by means of throughput & probability

[Bob Copeland <me@bobcopeland.com>]

On Sat, Mar 02, 2013 at 12:01:59PM +0100, Thomas Hühn wrote:
> Hi Bob,
> 
> I should have removed the debugging, thx for finding, I will fix this in v2.

Thanks!

> If I understand you right, your idea is to avoid equal rates for
> max_throughput and max_probability. The patch above would use equal
> rates for max_thr & max_prob. if the channel conditions allow this, so
> if they are pretty good. In other cases the max. probability rate can
> be totally different from max_thr. rate. So I do not understand the
> reasoning behind enforcing different rates for max_tp and max_prob.
> Minstrel will just apply its rule of max_prob_rate regardless if this
> rate also provides best or second best or other throughput values.

I guess my reasoning is that if you have an MRR setup like this:

mrr0 rate = 24, count = 8  (max_tp)
mrr1 rate = 18, count = 3  (max_tp2)
mrr2 rate = 24, count = 8  (max_prob)
mrr3 rate = 6,  count = 3

...and you are located right next to the microwave that just got
turned on :) ... then that second set of 24 mbit retries may well be
quite some time of wasted airtime vs perhaps picking 12 mbit if the
probability of those rates is close.

[this was an actual table generated by minstrel from my simulation,
at least at the time, max retries was ignored by minstrel.]

> Would you mind to this this patch with your hwsim ?

That is a bit complicated :)  To do so requires porting my userspace
wifi simulator to wmediumd, a process I have started but haven't
put much time into.  Nevertheless it has been on my personal to-do for
quite some time, so I'll give it a go soon.  I wouldn't hold up
acceptance of the patches for my testing though.

-- 
Bob Copeland %% www.bobcopeland.com