[RFC 0/4] hostapd: introduce Automatic Channel Selection (ACS)

[Yogesh Ashok Powar]

On Fri, Jul 26, 2013 at 02:13:04AM -0700, Michal Kazior wrote:
> On 26 July 2013 10:45, Yogesh Ashok Powar <yogeshp at marvell.com> wrote:
> > Hi Michal,
> >
> > Here is our last findings from ACS feature in hostapd
> > http://article.gmane.org/gmane.linux.kernel.wireless.general/109945
> >
> > We can give users some option to choose from different ACS
> > algorithms. Eg., Simple scan based algorithm or Advanced algo which
> > requires off-channel rx/tx support from the driver.
> 
> Hi,
> 
> Thanks! I've missed the thread apparently.
> 
> My patchset addresses both issues a) and b) you mentioned.
> chan_time_tx is not required now (it's used if it's available) and it
> also falls back to use chan_time_rx instead of chan_time_busy (ath10k
> is capable of reporting channel_time, channel_time_rx and noise floor
> only; no channel_time_tx/rx).
Great.

> 
> I've been wondering about parametrizing the noise floor treshold
> multiplier that's used to prefer channels with lower noise floor in
> case there's very little to no channel load. I'm unsure about the
> current threshold between base noise floor channel versus channel load
> though.
Sounds interesting. We faced issues with noise flooring since small
error in noise calculation will have larger effect on overall
interference factor because its power of 2. NF threshold mulitplier
might solve this issue.


BTW, we have written a simple scan based algo. We have sanity tested
that on 2.4 GHz only. For quick view, take a look at http://ix.io/6SA .
Open for optimization.

Thanks
Yogesh
> 
> 
> Pozdrawiam / Best regards,
> Micha? Kazior.