Re: [PATCH net-next 2/7] tcp: track min RTT using windowed min-filter

[Andrew Shewmaker <agshew@gmail.com>]

On Wed, Oct 14, 2015 at 02:28:00AM -0700, Andrew Shewmaker wrote:
> On Fri, Oct 16, 2015 at 09:57:42PM -0700, Yuchung Cheng wrote:
> ...
> > diff --git a/include/linux/tcp.h b/include/linux/tcp.h
> > index 86a7eda..90edef5 100644
> > --- a/include/linux/tcp.h
> > +++ b/include/linux/tcp.h
> > @@ -217,6 +217,9 @@ struct tcp_sock {
> >  	u32	mdev_max_us;	/* maximal mdev for the last rtt period	*/
> >  	u32	rttvar_us;	/* smoothed mdev_max			*/
> >  	u32	rtt_seq;	/* sequence number to update rttvar	*/
> > +	struct rtt_meas {
> > +		u32 rtt, ts;	/* RTT in usec and sampling time in jiffies. */
> > +	} rtt_min[3];
> 
> First, thanks for all the work in this patch series. In particular,
> applying Kern's check to ca_seq_rtt_us should fix some bad behavior
> I've observed recently.

I'd have to run more tests to be sure, but net-next with this patch series
significantly improves upon the poor behavior I was seeing where randomly
dropped packets greatly limited the usefulness of RTTs as a signal of
congestion. I still see some measurements that exceed the amount of delay
possible from queue buildup, but fewer than before.

Thanks!

-Andrew