.
2: Assign all multicast traffic to a single flow ID (eg. zero), without
reassigning the tin. This will cause it all to be treated like a single flow,
giving the FQ mechanisms something to bite on.
- Jonathan Morton
___
Bloat mailing list
> On 21 May, 2024, at 8:32 pm, Sebastian Moeller wrote:
>
>> On 21. May 2024, at 19:13, Livingood, Jason via Bloat
>> wrote:
>>
>> On 5/21/24, 12:19, "Bloat on behalf of Jonathan Morton via Bloat wrote:
>>
>>> Notice in particular
pens when
L4S traffic meets a conventional AQM.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
so that in practice all flows are soon
seen as "old". The DRR++ mechanism doesn't suffice, because the state in
Confucius is supposed to evolve over longer time periods, much longer than the
sojourn time of an individual packet in the queue.
The basic idea is interesting, bu
recisely the same phenomenon of short-term
queuing, but observe it in the form of the limited delivery rate of a burst,
rather than an increase in delay on the later packets of the burst.
- Jonathan Morton
___
Bloat mailing list
Bloat
there's one we can recommend with the "new wifi stack", so much the better.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
net users
to move to a truly different service, especially one based on a different link
technology. When attempts are made to increase competition, for example by
deploying a publicly-funded network, the incumbents actively sabotage those
attempts by any means they can. Monopolies are inherently cus
) DSCP field.
Apparently Comcast just found a whole raft of these in their own network as
part of rolling out L4S support. Funny how they didn't notice them previously.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferblo
> On 26 Aug, 2023, at 2:48 pm, Sebastian Moeller via Ecn-sane
> wrote:
>
> percentage of packets marked: 100 * (2346329 / 3259777) = 72%
>
> This seems like too high a marking rate to me. I would naively expect that a
> flow on getting a mark scale back by its cwin by 20-50% and then slowly
of TCP flags
and ECN field values depending on whether RFC-3168 or AccECN is being
attempted. Without AccECN, you won't have functioning L4S on a TCP stream.
But I think it is more likely that it's a misapplied DSCP.
- Jonathan Morton
___
Bloat
onvenient.
There are methods, however, of observing how available capacity tends to change
over time (typically on diurnal and weekly patterns, if the variations are due
to congestion in the ISP backhaul or peering) and scheduling adjustments on
that basis. If you have more information on y
ldn't even know to
consider such a failure mode, let alone be able to infer what they could do to
mitigate it.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
at, due to new-flow effects, the effective utilisation of
the path is only 50% over those two seconds. A longer flow might have better
utilisation in its later stages.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lis
rs are now
learning the value of implementing AQM (to apply congestion signals explicitly,
before the buffers are full), or failing that, of sizing the buffers
appropriately so that path latency doesn't increase unreasonably before
congestion signals are natu
luence QoE will you succeed.
Unfortunately, many QoS schemes are inadequate for the needs of actual Internet
users; this is because their designers have not kept up with the appropriate
QoE factors.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
ptimum of network power, using
information from the network to guide endpoint behaviour. This is *only*
possible using explicit information from the network, however, and is not
directly compatible with the current congestion-control paradigm of
RTT-fairness by default.
- Jonathan M
-end delay measurements will
> occasionally get it wrong and this can lead to starvation". Seems related to
> Jaffe's work on network power (titled "Flow control power is
> non-decentralizable").
Hasn't this been known for many years, as a consequence of
s
implications for setting AQM targets and tolerances in even near-ideal network
environments. But they did also note that BBR showed much less sensitivity to
this effect, as it uses pacing. In any case, it confirms that the first role
of a buffer is to absorb bursts with
in particular,
> seeing tcp tested more than unicast udp has been, in more labs, has long been
> on my mind.
Yes, as a tool specifically for testing with, and distributed with copious
warnings against attempting to use it more generally, this might be interesting.
- Jonathan Morton
_
ne that
nothing visibly changed.
At any rate, the original problem turned out to be something else entirely.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
t/69341657
A browser-based speed test like DSLreports depends heavily on the
responsiveness of the browser itself. It would appear that RDP interferes with
that quite spectacularly, although I'm unsure exactly why. The only advice I
can give is "don't do that, then".
- Jonathan Mort
nk would be reflected in that, such as the early rollout of fq_codel by
free.fr.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
> On 9 Aug, 2021, at 10:25 pm, Dave Collier-Brown
> wrote:
>
> My run of it reported latency, but without any qualifiers...
One would reasonable assume that's idle latency, then.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.buf
o
> each other, and the packet rate drops through the floor until they stop
> having relative motion. And I assume that also applies to time-varying
> path-loss and path-distance (multipath reflections).
So is it time to mount test stations on model railway wag
On Wed, 4 Aug 2021 at 21:31, Juliusz Chroboczek wrote:
> A Cortex-A53 SoC at 1GHz with correctly designed Ethernet (i.e. not the
> Raspberry Pi) can push 1Gbit from userspace without breaking a sweat.
That was true of the earlier Raspberry Pis (eg. the Pi 3 uses a brace
of Cortex-A53s) which use
USB Ethernet and Wifi
hardware plugged into it. Either will do the job withhout any
Ethernet bottlenecks, although the capabilities of USB Wifi dongles
are usually quite limited.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https
ould also be implemented in the MAC, but I don't know of
any vendor that's done that yet.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
ted.
Increased levels of sophistication in both the AQM and the endpoint's
congestion control algorithm may be used to increase the "network power"
actually obtained. The required level of complexity for each, achieving
reasonably good results, is however quite low.
- Jonathan M
It is "Rounds Per Minute", Apple's new measure
of network responsiveness.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
riterion 2 being
false. The number of flows going to even a family household is probably in the
low dozens at best. A control-theory approach can also work here.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
ding with limiting the amount of inflight data
So this corresponds to approach a) in Roland's taxonomy.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
em is not entirely petty - it is the difference of a
line item on the monthly bill. I'm sure you can see how the perverse
incentives arise with that.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
s those
determine what effect a bulk application load has on other, more
latency-sensitive applications. The two are also optimally controlled by
different mechanisms - FQ versus AQM - which is why the combination of the two
is so powerful.
Feel free to use material from the above with approp
are not lost, the flows experiencing them are not disadvantaged and the
so-called "capture effect" will not occur.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
> On 11 Jun, 2021, at 10:14 pm, Nathan Owens wrote:
>
> round-trips per minute
Wow, one of my suggestions finally got some traction.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
ttleneck by implementing a
faster connection to the NIC that can support wire-speed transfers.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
rstanding, and maybe filled in later. Things like
Diffserv, the URG pointer, option fields, and socket timeouts are not relevant
topics. There's no need to actually hide them from a header diagram, but just
highlight the fields that are fundamental to getting a payl
hat implementing AQM well is a key enabler towards those improvements.
That is probably the right perspective for "selling" AQM to them.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
y implements AF-AQM
commercially. My understanding is that similar functionality can also be added
to many recent cable head-ends by a firmware upgrade.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
lay to the
bulk traffic flows. I would suggest that if you implement FQ, you can also
usually implement AQM on top with little difficulty.
Please do ask for further clarification if that would be helpful.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
mplification…
I suspect such considerations are well beyond the level of education requested
here. I think what was being asked for was "how do these protocols work, and
why do they work that way, in language suitable for people working in a
different field", rather than "which one should
is well.
It's common to use TCP for transferring files or establishing a persistent
command-and-control connection. It's common to use UDP for simple
request-response applications (where both the request and response are small)
and where timeliness of delivery is far more important than relia
> On 13 May, 2021, at 12:10 am, Michael Richardson wrote:
>
> But, I'm looking for terminology that I can use with my mother-in-law.
Here's a slide I used a while ago, which seems to be relevant here:
The important thing about the term "quick" in this context is that throughput
capacity can
> On 17 May, 2021, at 8:18 am, Simon Barber wrote:
>
> How’s that?
It's a wall of text full of technical jargon. It seems to be technically
correct, but probably not very useful for the intended context.
- Jonathan Morton
___
Bloat mai
> On 17 May, 2021, at 12:33 am, Jonathan Morton wrote:
>
> The delay is caused by the fact that the product already in the pipeline has
> already been bought by the hardware store, and thus contractually the loggers
> can't divert it to an individual customer like me.
The r
then I might
only have to wait for the throughput of of the operation to produce what I
needed, and load it directly into my trailer. *That* would be just-in-time
manufacturing.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferblo
es waiting for trimming.
I planned for half an hour. It actually took me three hours to get my
firewood. Not for lack of throughput - that was one pretty effective logging
operation - but because of the *queues*.
- Jonathan Morton
___
Bloat mailing
ty valve.
Overall, it's very much a hybrid approach.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
> On 9 Mar, 2021, at 10:38 pm, Dave Taht wrote:
>
> sudo sysctl -w net.inet.tcp.disable_tcp_heuristics=1
Now that might well be the missing link. I think we missed it before since it
doesn't have "ecn" in its name.
- Jonathan Morton
__
This is actually something Cake's shaper can already cope with. I did some
testing on an ancient PC that didn't have HPET hardware, so timer interrupts
only had 1ms resolution even on Linux. This merely results in small bursts of
traffic at 1ms intervals, which collectively add up to the configured
s
very much not the case some years ago). However, there's no tariff at any
convenient level between 1Mbps (poverty tariff) and 50Mbps (probably radio
limited on a single carrier).
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
he inherent path latency is (reportedly) similar to a
terrestrial path and much shorter than a geostationary bounce.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
of 10G ports are part of the switch
portion, and thus intended to support LAN rather than WAN traffic. Think of it
as equivalent to attaching a Raspberry Pi 4 (which has native GigE) to a switch
with a pair of 10G "uplink" ports for daisy-chaining to oth
l Cake on it, if it isn't there already. Please give it a try and let us
know if it improves matters.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
> On 8 Jan, 2021, at 5:38 pm, Neal Cardwell via Make-wifi-fast
> wrote:
>
> What did you have in mind by "variable links" here? (I did not see that term
> in the paper.)
Wifi and LTE tend to vary their link characteristics a lot over
n the consumer-end modem.
This is fixable, just as soon as Starlink put their minds to it, because it's
based on the same Atheros SoCs as the consumer CPE we're already familiar with.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.buf
aturating ECN Capable flow through it and looking for
CE marks (and/or ECE feedback), since Airtime Fairness comes with built-in
fq_codel.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
itself. Distinct upload and download bloat tests would help in determining
whether it's at your end or the remote end. You should be able to use
dslreports.com/speedtest to determine that.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.
download direction), you need to redirect the
ingress traffic to an IFB device and attach an ingress-configured Cake instance
there. You would use "ingress" instead of "ack-filter" and your download
bandwidth.
For egress traffic you should indeed use the upload speed.
- Jonathan
s.
For your topology, eth0 (LAN egress) should get dual-dsthost, and eth1 (WAN
egress) should get dual-srchost.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
slow and medium-speed
broadband (up to 100Mbps), and can also be used at higher speeds with some
care, mostly regarding device performance.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
s a lot of kernel time spent waiting for the hardware, and can
only really be solved by redesigning the hardware.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
towards, though I suspect that the way Cake's shaper
is integrated is still better than having an external one in software.
With that said, it's also possible that something a bit lighter than Cake might
be appropriate at cable speeds. There is background work in this general area
going on,
might be 10% of what the cable company is advertising
to them.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
least
allow running ingress and egress on different CPUs.
Is there an example anywhere on how to do this?
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
.
There are no easy answers here. But I've suggested some things to look
for and try out.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
On Tuesday, 18 August 2020, Daniel Sterling wrote:
> As you know, I'm here cuz I have an xbox and y'all created cake, which
> I am eternally grateful for, since it makes latency go away.
>
> But I've recently hit an interesting issue --
>
> Microsoft (and/or akamai, or whatever) has recently
> The current best practice seems to be to instantiate cake/SQM on a reasonably
> fixed rate wan link and select WiFi cards/socs that offer decent airtime
> fairness.
> Works pretty well in practice...
Yes, AQL does essentially the right thing here, again along the lines of
limiting the
> Are the risks and tradeoffs well enough understood (and visible enough
> for troubleshooting) to recommend broader deployment?
>
> I recently gave openwrt a try on some hardware that I ultimately
> concluded was insufficient for the job. Fairly soon after changing out
> my access point, I
is the
latter which inspired the name.
For the other major functional component of fq_codel, Cake also has a
set-associative hash function for allocating flows into queues, which
substantially reduces the probability of hash collisions in most cases.
- Jonathan Morton
ditioning of the data it gathers about
the network path.
I'm not altogether a fan of such complexity.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
windows from
causing a huge burst which would tend to swamp buffers.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
re reasonable to implement a P2P communication strategy for a PvE
game. The central server is then only responsible for coordinating enemy
movements.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
n without adding any load at all. There is also
connmon (https://github.com/pollere/connmon).
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
> On 3 Jun, 2020, at 7:48 pm, Dave Taht wrote:
>
> I am of course, always interested in how they are measuring latency, and
> where.
They don't seem to be adding more latency measurements once the download tests
begin. So in effect they are only measuring idle latency.
- Jon
dered quite a heavyweight solution, but very effective. If it
doesn't work well for this particular use case, it may be feasible to backport
some more recent work which takes a simpler approach, though along similar
lines.
- Jonathan Morton
__
ng be on the usb0 device and/or the tunX?
You should set the qdisc and those options on the *physical* device, not the
one that carries your uncompressed data. Don't forget to set up ingress
shaping as well as egress.
- Jonathan Morton
___
iptions. Roughly €25 will buy you
500Kbps mobile service for three months, and you can use that 500Kbps as much
as you like. And that is with the lowest population density in Europe, so the
per capita cost of covering the country in cell towers is obviously no excuse
links,
because with the built-in shaper it has knowledge of the speed. I don't think
I've tested it as low as 16Kbit, but I have used it at 64Kbit.
To keep things simple, you may want to specify "besteffort flows satellite" as
parameters. Some of those settings may also be
which expects and can handle it. The SCE proposal also
inserts AF or FQ protection at these nodes, which serves as a prophylactic
against the likes of DCTCP being used inappropriately on the Internet.
- Jonathan Morton
___
Bloat mail
n the room understood
that.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
sponse at their end is definitely not Multiplicative Decrease.
I haven't dug into it more deeply than that. But they're also not running
AccECN, nor are they "proactively" sending CWR to get a "more accurate" CE
feedback. I suspect t
be that your version of fq_codel doesn't actually have ECN
support on by default. So try adding the "ecn" keyword to the qdisc.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
OSX:
$ sudo sysctl -w net.inet.tcp.ecn_initiate_out=1
$ sudo sysctl -w net.inet.tcp.ecn_negotiate_in=1
In Linux and OSX, to make the setting persist across reboots, edit
/etc/sysctl.conf.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists
t; (The link is supposed to be 50 Mbps symmetric and speed test does show it
> bursting that high sometimes.)
Looks like a definite improvement. The Quality grade of C may indicate that
you haven't enabled ECN on your client; without it, Codel has to drop packet
efinitely too high; you don't have complete
control of the queue here, and that's visible particularly with the steady
increase in the upload latency during the test. Try 44500 up, 42000 down,
equivalent to my suggestions for Cake.
- Jonathan Morton
___
t more flows than another.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
bsockets error during the test, so try it again and
it might work.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
Teleconferencing, by contrast, tends to involve multiple
audio and video streams going everywhere.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
ifi hardware. Otherwise the benefit will only be
limited. AQM and/or FQ has to be applied at the bottleneck; sometimes a
bottleneck has to be artificially induced to implement that.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferblo
it set itself Not-ECT?
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
streams based on UDP/TCP port number,
This is the ideal situation for simply deploying Cake without any special
effort. Just tell it the capacity of the link it's controlling, minus a modest
margin (say 1% upstream, 5% downstream).
You should be pleasantly surprised by the results
t; enabling technology. Just as we suspected.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
to set the extra set of Codel parameters
involved.
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
k is never delayed just to see if it can be combined with
a later one. The result is a better use of limited capacity to carry useful
payloads, without having to rely on dropping acks by AQM action (which Codel is
actually rather bad at).
- Jonathan Morton
__
erved
bits are covered yet, only the one we use).
- Jonathan Morton
___
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
extremes of 100% relative error in the feedback,
which is shown to be safe and reasonably tolerable. Smaller errors due to
random ack loss are more likely, and consequently easier to tolerate in a
closed negative-feedback control loop.
- Jonathan Morton
__
ut unknown TCP options
tend to get either stripped or blocked.
This influenced the design of AccECN as well; in an early version it would have
used only a TCP option and left the TCP flags alone. When it was found that
firewalls would often interfere with this, the three-bit field in the
deployment in practice in recent years. He invited me, one of the ICCRG
chairs, and Bob Briscoe - among others - to dinner, where we discussed some
technical distinctions and Bob demonstrated a fundamental misunderstanding of
control theory.
And we will have more ammunition at Vancouver
eue lengths proposed by
L4S, the example they give in the Prague Requirements is of a 100ms versus 1ms
baseline path, lengthened to 101ms versus 2ms by a 1ms queue. This reduces a
100:1 ratio to 50.5:1.
The FQ example is, however, of the network enforcing fairness, rather than
infor
1 - 100 of 416 matches
Mail list logo