[Cake] [Bloat] some mikrotik comments

Sun Oct 10 20:40:22 EDT 2021

On Sun, Oct 10, 2021 at 4:01 PM Jordan Szuch <jordan at inacomptc.com> wrote:
>
>
>
> -----Original Message-----
> From: Dave Taht <dave.taht at gmail.com>
> Sent: Sunday, October 10, 2021 5:38 PM
> To: Jordan Szuch <jordan at inacomptc.com>
> Cc: Cake List <cake at lists.bufferbloat.net>; bloat <bloat at lists.bufferbloat.net>
> Subject: Re: [Bloat] some mikrotik comments
>
> Great!
>
> >Encouraged!
>
> On Sun, Oct 10, 2021 at 1:39 PM Jordan Szuch <jordan at inacomptc.com> wrote:
> >
> > Unfortunately, didn't have time to setup my own server but here are
> > the results for Flent on against the public east/west bufferbloat
> > servers: https://1drv.ms/u/s!Ap4u4Rte63FqjqQ8zMdwTTBmcd_mnw?e=675gyh
> >
> > That's a shared OneDrive folder with results from running with CAKE and then without any kind of shaping. Here's some additional information as well.
> >
> > This was the command that was run: flent rrul -p all_scaled -l 60 -H
> > netperf.bufferbloat.net
>
> >-t the_test_parameters # helps to name the plots, so you can compare them later
> >--step-size=.05 # sample at a higher frequency
>
> Got it. There's now a new run in the shared CAKE folder with the high sample frequency (Run #2)

The upload background flow is really struggling at this speed (5% of
the total bandwidth) but... this is probably a good result.

I am comforted that mikrotik is shipping a correct cake.

If you had 2x more uplink bandwidth, the observed delay on the
measurement flows would be closer to 0.

>
> > Internet is a Spectrum cable connection. Bandwidth is supposed to be 100Mbps down and 10Mbps up. I have it limited to 90Mbps and 9Mbps for shaping. Here's also the CAKE settings I have configured on the router:
>
> >Is it not a docsis 3.1 modem? Because your unshaped result was *miserable*.
>
> No, just 3.0. Most of the modems in this area are that presently (Northern Michigan). No much incentive for Spectrum to replace them unless you're purchasing their highest speeds. Not sure if a 3.1 modem would make a difference in their case, but it's possible.

At least on comcast now, all 3.1 modems have pie, and so do the
cmts-es. Rather than 1000ms+ delay you are seeing on spectrum, you
should see no more than 60ms, with pie on in both directions, on this
test. The fairness vs the cake result will be less, also, but it is
very hard to argue with a 16x1 reduction in latency under this load in
the out of the comcast-box experience.

I am certainly hoping that the rest of the cable industry follows
comcast, soon.

Someday perhaps they too will adopt cake, but I think they'd have to
be forced into it by competitive pressure from a fiber
provider, or a starlink.

>
> >
> > Upload Queue
> > name="cake-nat-doc-up" kind=cake cake-bandwidth=0bps cake-overhead=18
> >      cake-mpu=64 cake-overhead-scheme=docsis cake-rtt=100ms
> >      cake-diffserv=diffserv4 cake-flowmode=triple-isolate cake-nat=yes
> >      cake-wash=no cake-ack-filter=filter
>
> >0bps ?
>
> >with docsis framing you should be able to just say 10mbit, rather than 9.
>
> >cake-ack-filter=filter ? does it have filter-aggressive?
>
> A quirk of how you can configure the queue on the router. You can specify the bandwidth limit in the general queue configuration or the configuration for the queue type you're using. Configuration now looks something like this:
>
> name="cake-nat-doc-up" kind=cake cake-bandwidth=10.0Mbps cake-overhead=18
>      cake-mpu=64 cake-overhead-scheme=docsis cake-rtt=100ms
>      cake-diffserv=diffserv4 cake-flowmode=triple-isolate cake-nat=yes
>      cake-wash=no cake-ack-filter=aggressive
>
> name="cake-nat-doc" kind=cake cake-bandwidth=110.0Mbps cake-overhead=18
>      cake-mpu=64 cake-overhead-scheme=docsis cake-rtt=100ms
>      cake-diffserv=diffserv4 cake-flowmode=triple-isolate cake-nat=yes
>      cake-wash=no cake-ack-filter=none
>
> And yes, there is the aggressive filter.

I am in general unfond of the aggressive option, but ... pretty good.
It shouldn't be quite this variable. Go back to just filter.

If you want, try enabling ecn on the client.
> >
> > Download Queue
> > name="cake-nat-doc" kind=cake cake-bandwidth=0bps cake-overhead=18
> >      cake-mpu=64 cake-overhead-scheme=docsis cake-rtt=100ms
> >      cake-diffserv=diffserv4 cake-flowmode=triple-isolate cake-nat=yes
> >      cake-wash=no cake-ack-filter=none
>
> >It looks like your provisioned rate was closer to 120. Try 110.
>
> I increased the bandwidth as seen above. We'll see how that behaves!

LGTM. 105 perhaps. Or 115. Try it during netflix hours.

>
> > Hopefully those settings are configured somewhat decently. Hope that helps!
> >
> > Jordan
> >
> > -----Original Message-----
> > From: Dave Taht <dave.taht at gmail.com>
> > Sent: Sunday, October 10, 2021 1:51 PM
> > To: Jordan Szuch <jordan at inacomptc.com>
> > Cc: Cake List <cake at lists.bufferbloat.net>; bloat
> > <bloat at lists.bufferbloat.net>
> > Subject: Re: [Bloat] some mikrotik comments
> >
> > goferit! We have flent servers throughout the world, mostly undocumented. I tend to prefer folk setup their own, less noise in the data....
> >
> > On Sun, Oct 10, 2021 at 9:18 AM Jordan Szuch <jordan at inacomptc.com> wrote:
> > >
> > > OK, I'll give it a shot! Please forgive me here as I haven't really run flent before. Would running something like what's listed in the Quick Start here (https://flent.org/intro.html#quick-start) be sufficient?
> > >
> > > Jordan
> > >
> > > -----Original Message-----
> > > From: Dave Taht <dave.taht at gmail.com>
> > > Sent: Sunday, October 10, 2021 12:11 PM
> > > To: Jordan Szuch <jordan at inacomptc.com>
> > > Cc: Cake List <cake at lists.bufferbloat.net>; bloat
> > > <bloat at lists.bufferbloat.net>
> > > Subject: Re: [Bloat] some mikrotik comments
> > >
> > > It's flent.org rrul test output that I love.
> > >
> > > On Sun, Oct 10, 2021 at 9:05 AM Jordan Szuch via Bloat <bloat at lists.bufferbloat.net> wrote:
> > > >
> > > > Hi Dave,
> > > >
> > > > Interesting that you should send this out; I recently picked up a unit (this one specifically: https://mikrotik.com/product/hap_ac2) to do some testing of the v7 firmware. Nothing super high end but enough to do a little testing. I have CAKE set to do some shaping and so far seems functional. At least the bufferbloat speed tests from DSLReports and WaveForm indicate some shaping is happening. If you'd like any information about the unit, the v7 firmware, or particular tests run then I'd be happy to help out.
> > > >
> > > > Jordan
> > > >
> > > > -----Original Message-----
> > > > From: Bloat <bloat-bounces at lists.bufferbloat.net> On Behalf Of
> > > > Dave Taht
> > > > Sent: Sunday, October 10, 2021 11:39 AM
> > > > To: Cake List <cake at lists.bufferbloat.net>; bloat
> > > > <bloat at lists.bufferbloat.net>
> > > > Subject: [Bloat] some mikrotik comments
> > > >
> > > > the v7 beta gained doc and more os support for fq_codel and cake
> > > > recently
> > > >
> > > > https://forum.mikrotik.com/viewtopic.php?p=885000#p885000
> > > >
> > > > anyone out there actively testing mikrotik?
> > > >
> > > > --
> > > > Fixing Starlink's Latencies:
> > > > https://www.youtube.com/watch?v=c9gLo6Xrwgw
> > > >
> > > > Dave Täht CEO, TekLibre, LLC
> > > > _______________________________________________
> > > > Bloat mailing list
> > > > Bloat at lists.bufferbloat.net
> > > > https://lists.bufferbloat.net/listinfo/bloat
> > > >
> > > > _______________________________________________
> > > > Bloat mailing list
> > > > Bloat at lists.bufferbloat.net
> > > > https://lists.bufferbloat.net/listinfo/bloat
> > >
> > >
> > >
> > > --
> > > Fixing Starlink's Latencies:
> > > https://www.youtube.com/watch?v=c9gLo6Xrwgw
> > >
> > > Dave Täht CEO, TekLibre, LLC
> > >
> >
> >
> > --
> > Fixing Starlink's Latencies:
> > https://www.youtube.com/watch?v=c9gLo6Xrwgw
> >
> > Dave Täht CEO, TekLibre, LLC
> >
>
>
> --
> Fixing Starlink's Latencies: https://www.youtube.com/watch?v=c9gLo6Xrwgw
>
> Dave Täht CEO, TekLibre, LLC
>

-- 
Fixing Starlink's Latencies: https://www.youtube.com/watch?v=c9gLo6Xrwgw

Dave Täht CEO, TekLibre, LLC