[Cake] [PATCH net-next] sched: sch_cake: Align QoS treatment to Windows and Zoom

[Dave Taht]

actually in re-reading that, I think it leaves nqb in 0 (44, not 45,
is now VI), which wasn't my intent.

More below...

On Fri, Jan 10, 2025 at 8:24 AM Sebastian Moeller <moeller0 at gmx.de> wrote:
>
> Mmmh. off the Linux lists/folks for discussion:
>
> can I propose to drop NQB down to BestEffort? The supposed requirements for NQB marking should make these flows thrive in a flow isolating scheduler like cake without any special treatment... moving it to Video or worse Voice really only invites abuse...

I would prefer consistent behavior.

>
>
>
> > On 10. Jan 2025, at 16:55, Toke Høiland-Jørgensen via Cake <cake at lists.bufferbloat.net> wrote:
> >
> > From: Dave Taht <dave.taht at gmail.com>
> >
> > Cake's diffserv4 mode attempted to follow the IETF webrtc
> > QoS marking standards, RFC8837.
> >
> > It turns out Windows QoS can only use CS0, CS1, CS5, and CS7.
>
> Nah, applications apparently can only request a small set of DSCPs, but it is possible to make whole applications use a different DSCP:
> example for making putty use EF
>
> New-NetQosPolicy -Name "putty" -AppPathNameMatchCondition "putty.exe" -PolicyStore ActiveStore -NetworkProfile All -DSCPAction 46

Within an App...

>
>
> >
> > Zoom defaults to using CS5 for video and screen sharing traffic.
>
> Zoom defaults to using CS0 (my packet captures) unless an adminstrator enables DSCPs, which I assume will only/mostly happen for business accounts.

I have seen it enabled. I have also multiple reports of "stuttering",
which I am hoping this resolves.

>
> > Bump CS4, CS5, and NQB to the video tin (2) in diffserv4 mode, for
> > more bandwidth and lower priority.
>
> I do endorse that, but that change will affect all current users of cake as parts of their individual QoS hierarchies.

I think windows users are not as  clued.
> >
> > This also better aligns with how WiFi presently treats CS5 and NQB.
>
> Two issues: it is not that default WiFi uses a sane system to begin with, and NQB IMHO should not be elevated in priority unless it is also policed well, something neither WiFi nor cake will do out of the box.
>
>
> >
> > Signed-off-by: Dave Taht <dave.taht at gmail.com>
> > Signed-off-by: Toke Høiland-Jørgensen <toke at redhat.com>
> > ---
> > net/sched/sch_cake.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
> > index 48dd8c88903f..2a9288d4b873 100644
> > --- a/net/sched/sch_cake.c
> > +++ b/net/sched/sch_cake.c
> > @@ -328,8 +328,8 @@ static const u8 diffserv4[] = {
> > 1, 0, 0, 0, 0, 0, 0, 0,
> > 2, 0, 2, 0, 2, 0, 2, 0,
> > 2, 0, 2, 0, 2, 0, 2, 0,
> > - 3, 0, 2, 0, 2, 0, 2, 0,
> > - 3, 0, 0, 0, 3, 0, 3, 0,
> > + 2, 0, 2, 0, 2, 0, 2, 0,
> > + 2, 0, 0, 0, 2, 0, 3, 0,
> > 3, 0, 0, 0, 0, 0, 0, 0,
> > 3, 0, 0, 0, 0, 0, 0, 0,
> > };
> > --
> > 2.47.1
> >
> > _______________________________________________
> > Cake mailing list
> > Cake at lists.bufferbloat.net
> > https://lists.bufferbloat.net/listinfo/cake
>


-- 
Dave Täht CSO, LibreQos