CAKE w/ Adaptive Bandwidth

[riccy]

ok

here are the results:

with sqm and cakeautorate DISABLED

Dropbox

with sqm and cake autorate ENABLED

Dropbox

As mentioned before this test gives me the lowest speeds by far.

must have snuck in before your ip changed

[moeller0]

Disabled:

Enabled:

Disregarding the spikes, under load sqm+autorate seem to actually improve things with the exception of a few nasty latency spikes...

[moeller0]

Disabled, all:

Enabled, all:

[Lynx]

And this is all still with the bandwidths set way higher than they should be, right? I would love to see the cake plus cake-autorate tests repeated with more appropriate bandwidth settings. But I don’t know what they should be set to? Maybe min 25Mbps, base 50Mbps and max 300Mbps? And I already suggested reduced delay settings above:

@Lynx

For your connection I think the defaults might actually be too lax and I think that more aggressive settings might help better reduce latency:

# average owd delta threshold in ms up to which the maximum adjust_up_load_high is applied to the shaper rate adjustment
# for average owd deltas between avg_owd_delta_max_adjust_up_thr_ms and owd_delta_thr_ms, the adjustment is scaled linearly
# from max_adjust_up_load_high (at avg_owd_delta_max_adjust_up_thr_ms) to min_adjust_up_load_high (at owd_delta_thr_ms)
dl_avg_owd_delta_max_adjust_up_thr_ms=10.0 # (milliseconds)
ul_avg_owd_delta_max_adjust_up_thr_ms=10.0 # (milliseconds)

# owd delta threshold in ms is the extent of OWD increase to classify as a delay
# these are automatically adjusted based on maximum on the wire packet size
# (adjustment significant at sub 12Mbit/s rates, else negligible)
dl_owd_delta_delay_thr_ms=20.0 # (milliseconds)
ul_owd_delta_delay_thr_ms=20.0 # (milliseconds)

# average owd delta threshold in ms beyond which the maximum adjust_down_bufferbloat is applied to the shaper rate adjustment
# for average owd deltas between owd_delta_thr_ms and avg_owd_delta_max_adjust_up_thr_ms, the adjustment is scaled linearly
# from min_adjust_down_bufferbloat (at owd_delta_thr_ms) to min_adjust_up_load_high (at avg_owd_delta_max_adjust_down_thr_ms)
dl_avg_owd_delta_max_adjust_down_thr_ms=30.0 # (milliseconds)
ul_avg_owd_delta_max_adjust_down_thr_ms=30.0 # (milliseconds)

What can be deduced about @riccy’s connection from the data?

[riccy]

appreciate everyones views.

Will be back Wednesday- going for some r&r.

take care

keith

[moeller0]

So I would like to see LibreQoS tests with just sqm (so no autorate) set to 100/10, 200/20, 300/30. Preferably with concurrent irtt tests.
My question here is to what degree are the latency spikes we see in CAKE w/ Adaptive Bandwidth - #706 by moeller0 are sqm's doing?
And then whether we can remedy them to some degree? My hunch is these are not directly coming from cake dropping packets (the small green and red ticks at 0 of the x-axis denote dropped packets, albeit only from the sparse irtt traffic).
Side question @riccy are you testing via LAN cable or via WiFi?

@Lynx

I would love to see the cake plus cake-autorate tests repeated with more appropriate bandwidth settings. But I don’t know what they should be set to? Maybe min 25Mbps, base 50Mbps and max 300Mbps?

No idea on the numbers but I also think that the autorate settings could be tweaked a bit.

@Lynx

And I already suggested reduced delay settings above:

Unsure, there still is the oddity that his latency (even in the LibreQoS test seems to shrink noticeably during load in upstream direction... @Lynx you do not see something like that with your mobile-ISP and modem?

[riccy]

Hi guys,

Will try to do more tests when Im back.

By the, I am on lan, no wifi

[Lynx]

For a reference point, here is some data that @moeller0 plotted for me showing how my own 4G connection presently behaves with a few internet speed tests when cake is disabled:

Observe how the download latency delta is generally higher than the upload latency delta. On my connection, download bufferbloat is significantly worse.

@moeller0

Unsure, there still is the oddity that his latency (even in the LibreQoS test seems to shrink noticeably during load in upstream direction... @Lynx you do not see something like that with your mobile-ISP and modem?

Yes I think we also discovered earlier that the download latency delta on my 4G connection actually reduces with any active substantial upload.

I'll try to show this later with some more irtt data showing mixed upload and download using my modified betterspeedtest.sh script that supports bidirecitonal mode.

Actually @moeller0 this recent IRTT data looks rather different to back in April 2023:

CAKE w/ Adaptive Bandwidth [August 2022 to March 2024]

True, but I'm just looking for something quick. @richb-hanover-priv might you be tempted to add a bidirectional option to betterspeedtest.sh? It's not quite as simple as just: measure_direction "Download" & measure_direction " Upload" I'd need to look at it in detail to get that to work properly, but for now it's good enough to generate the loads. @moeller0 here are the files: Here is a screen capture of the loads reported by OpenWrt: [image] And excel plot of the irtt data: [image] …

Back then I noticed that bufferbloat seemed to be present mostly in the send direction. But that's not what we're seeing now, right?

The most recent plot seems to me to show mostly receive bufferbloat.

I wonder what is responsible for this difference. Perhaps some change in how traffic is handled at the cell tower?

[moeller0]

@Lynx

Back then I noticed that bufferbloat seemed to be present mostly in the send direction. But that's not what we're seeing now, right?

The most recent plot seems to me to show mostly receive bufferbloat.

I agree.

@Lynx

I wonder what is responsible for this difference. Perhaps some change in how traffic is handled at the cell tower?

Likely, that is no idea whether it is precisely at the tower, but likely inside the carrier's network...