CakeQOS CakeQOS-Merlin

[Pak Kriss]

Thanks for your feedback. For testing I switched to:

[1] --> Download Speed | [195 Mbit]
[2] --> Upload Speed | [195 Mbit]
[3] --> Queue Priority | [besteffort]
[4] --> Extra Download Options | [ethernet dual-dsthost]
[5] --> Extra Upload Options | [ethernet dual-srchost]

Just providing a datapoint, my parent's connection is cable at a modest 30/5, and the settings that seem to work best for them are:

Download: 28.5 (95% of speed)
Upload: 4.75 (95% of speed)
Queue: besteffort
Extra Download Options: docsis dual-dsthost
Extra Upload Options: docsis dual-srchost

I ran it a while without the dual-xhost entries (meaning it was running on the default triple-isolate). It did generally a good job but bufferbloat tests were a little inconsistent.

As mentioned on https://openwrt.org/docs/guide-user...pe_without_a_safety_net_aka_advanced_features:



If anyone else wants to test this out, there is no need to enter the "ingress" option as mentioned in this link as we already include it in our script for the "download" cake command.

Note: I don't use ack-filter as @Wade Coxon shared this article with me which suggests this option is good for 10+:1 dl/ul ratios: http://blog.cerowrt.org/post/ack_filtering/ (my parent's connection is 6:1). In other words, if your connection is something like 100/10 up/down, then you might be interested in testing out ack-filter.

EDIT: Sharing a screencap of the tests on DSLReports:

View attachment 24471

The tests up until mid-way to the 25.1 A+ result were on triple-isolate (you can see the results were inconsistent), everything after was with the above config.

Sent from my SM-T805 using Tapatalk

 

[rgnldo]

The presets option will be possible with the "0" unlimited option.
My example:
Preset: Fiber connection, bridge

 besteffort unlimited dual-dsthost DL "0"
 besteffort unlimited dual-srchost UP "0"

 

[Wade Coxon]

interesting
sounds like since cake-qos script applies wash to the download, diffserv4 might not work all that well?
it should either be "wash besteffort" or "nowash diffserv4" (unless you have comcast apparently).

"wash" as I understand it will just strip away any markings made by your ISP or other upstream providers before sending them to your router. That would be so that you can control the priorities yourself rather than rely on someone else's stab at what's best for you.

diffserv is pretty much useless at the moment as our cake implementation doesn't have any traffic marking of note. We just need a traffic identification/marking solution if we want something more elaborate than besteffort.

As has been said by others before, it would be good to be able to invoke the Trend engine for traffic classification only, and then have cake manage the traffic flow. I suspect that may be something that can't be done due to the way the packets need to flow through the Trend code blobs? (dunno, just speculating).

There was a link in the "I built cake" thread that included some discussion about traffic classification on openwrt. That looked like something promising that could be applied to cake here.
@SlotTech mentions it here
https://www.snbforums.com/threads/rt-ac86u-i-built-cake.49190/page-3#post-589090

I am having trouble finding the posted link in the thread, but I think it was pointing to this forum
https://forum.openwrt.org/t/ultimate-sqm-settings-layer-cake-dscp-marks/25832/107

There is a lot there and well above my pay-grade.

 

[Wade Coxon]

@SlotTech mentions it here
https://www.snbforums.com/threads/rt-ac86u-i-built-cake.49190/page-3#post-589090

I am having trouble finding the posted link in the thread, but I think it was pointing to this forum
https://forum.openwrt.org/t/ultimate-sqm-settings-layer-cake-dscp-marks/25832/107

There is a lot there and well above my pay-grade.

Actually found the link now.

It was @TechTinkerer in this very thread:
https://www.snbforums.com/threads/release-cakeqos-merlin.64800/page-20#post-596840

DSCP tagging on openwrt:
https://forum.openwrt.org/t/using-dscp-for-qos/14252/4

I am sure something along these lines would be useful to add to cake. Some basic tagging would go a long way to allowing us to meaningfully use one of the diffserv settings (I think 4 is plenty).

It seems like something that is going to put more load on the CPU, so those with fast connections may want the ability to stick with a tagless besteffort setup, or maybe limit the tagging to uploads only.

 

[ugandy]

good info. thanks
so, presently, "besteffort" is the right choice.

 

[Wade Coxon]

good info. thanks
so, presently, "besteffort" is the right choice.

So little traffic is classified at the moment, that I couldn't see there being any benefit at all in using anything else.

 

[Smokey613]

I keep getting this error while trying to install cake.


#########################################################
## _ ##
## | | ##
## ___ __ _ | | __ ___ __ _ ___ ___ ##
## / __|/ _` || |/ // _ \ ______ / _` | / _ \ / __| ##
## | (__ |(_| || <| __/|______| (_| || (_) |\__ \ ##
## \___|\__,_||_|\_\\___| \__, | \___/ |___/ ##
## | | ##
## |_| ##
## ##
## https://github.com/ttgapers/cakeqos-merlin ##
## v1.0.2 ##
## ##
#########################################################


#########################################################

Downloading http://bin.entware.net/aarch64-k3.10/Packages.gz
Updated list of available packages in /opt/var/opkg-lists/entware

#########################################################

CakeQOS-Merlin: Cake binaries missing - Exiting
Smokey613@RT-AC86U:/tmp/home/root#

 

[Wade Coxon]

Entware is probably still being unreliable.
It went down a couple of days ago. Note sure if it is still broken, or if it came back and re-broke .

 

[Smokey613]

I managed to remove all past traces of cake, ran the installer and got this.

 

[Adamm]

I managed to remove all past traces of cake, ran the installer and got this.

View attachment 24482

Manual intervention should not be required at all, thats probably why you are having issues downloading/installing the tc binary. Run the uninstall function first to properly clear your system.

skynet@RT-AX88U-DC28:/jffs/addons/cake-qos# ./cake-qos install

#########################################################
##               _                                     ##
##              | |                                    ##
##    ___  __ _ | | __ ___          __ _   ___   ___   ##
##   / __|/ _` || |/ // _ \ ______ / _` | / _ \ / __|  ##
##  | (__ |(_| ||   <|  __/|______| (_| || (_) |\__ \  ##
##   \___|\__,_||_|\_\\___|        \__, | \___/ |___/  ##
##                                    | |              ##
##                                    |_|              ##
##                                                     ##
##      https://github.com/ttgapers/cakeqos-merlin     ##
##                        v1.0.2                       ##
##                                                     ##
#########################################################


#########################################################

Downloading http://bin.entware.net/aarch64-k3.10/Packages.gz
Updated list of available packages in /opt/var/opkg-lists/entware
CakeQOS-Merlin: Updated sched-cake-oot_2020-05-28-a5dccfd8-ax_aarch64-3.10.ipk
No packages removed.
Installing sched-cake-oot (2020-05-28-a5dccfd8-ax) to root...
Configuring sched-cake-oot.
CakeQOS-Merlin: Updated tc-adv_4.16.0-git-20191110_aarch64-3.10.ipk
No packages removed.
Installing tc-adv (4.16.0-git-20191110) to root...
Installing libnl-tiny (2019-10-29-0219008c-1) to root...
Downloading http://bin.entware.net/aarch64-k3.10/libnl-tiny_2019-10-29-0219008c-1_aarch64-3.10.ipk
Installing libcap (2.33-2) to root...
Downloading http://bin.entware.net/aarch64-k3.10/libcap_2.33-2_aarch64-3.10.ipk
Installing libintl-full (0.19.8.1-2) to root...
Downloading http://bin.entware.net/aarch64-k3.10/libintl-full_0.19.8.1-2_aarch64-3.10.ipk
Installing libelf (0.179-1) to root...
Downloading http://bin.entware.net/aarch64-k3.10/libelf_0.179-1_aarch64-3.10.ipk
Installing libmnl (1.0.4-2) to root...
Downloading http://bin.entware.net/aarch64-k3.10/libmnl_1.0.4-2_aarch64-3.10.ipk
Configuring libcap.
Configuring libmnl.
Configuring libintl-full.
Configuring libnl-tiny.
Configuring libelf.
Configuring tc-adv.

#########################################################


Please enter your download speed:

[Mbit]: 92


Please enter your upload speed:

[Mbit]: 37


Select Queue Priority Type:
[1]  --> besteffort (default)
[2]  --> diffserv3
[3]  --> diffserv4
[4]  --> diffserv8

[1-4]: 1


Please enter your extra download options:
[Options]: ethernet


Please enter your extra upload options:
[Options]: ethernet


#########################################################

CakeQOS-Merlin: Starting - ( 92Mbit | 37Mbit | besteffort | ethernet | ethernet )
Broadcom Packet Flow Cache HW acceleration <DISABLED>
CPU Speed (*) ==> Disable
Changing power settings. Forcing cpu to max frequency.
   Set cpuspeed to on(pwr config --cpuspeed on) to make it adaptive again
Broadcom Packet Flow Cache learning via BLOG disabled.
Broadcom Packet Flow Cache flushing the flows

#########################################################

 

[Smokey613]

That is what I meant by removing past traces was I finally managed to run the uninstall function, then attempted a fresh install and still get these same errors.

 

[rgnldo]

Peformance in samba?
75 gb in 42 minutes with wifi?
Is it the effect of Cake-QOS?

 

[Adamm]

That is what I meant by removing past traces was I finally managed to run the uninstall function, then attempted a fresh install and still get these same errors.

Run this install command instead for additional output;

mkdir -p /jffs/addons/cake-qos && /usr/sbin/curl -s "https://raw.githubusercontent.com/ttgapers/cakeqos-merlin/master/cake-qos.sh" -o "/jffs/addons/cake-qos/cake-qos" && chmod 755 /jffs/addons/cake-qos/cake-qos && sh -x /jffs/addons/cake-qos/cake-qos install

Peformance in samba?
75 gb in 42 minutes with wifi?
Is it the effect of Cake-QOS?


View attachment 24489

23MB/s is relatively slow, with my NAS I can easily saturate a gigabit connection (113MB/s) over Wifi.

 

[brummygit]

interesting
sounds like since cake-qos script applies wash to the download, diffserv4 might not work all that well?
it should either be "wash besteffort" or "nowash diffserv4" (unless you have comcast apparently).

Exactly my reason for wondering why the decision was taken to apply wash to the download path. It could just be that whoever was doing the initial development had a comcast service.

I've seen no negative impact of removing it in my own copy of the script.

 

[brummygit]

diffserv is pretty much useless at the moment as our cake implementation doesn't have any traffic marking of note. We just need a traffic identification/marking solution if we want something more elaborate than besteffort.

As has been said by others before, it would be good to be able to invoke the Trend engine for traffic classification only, and then have cake manage the traffic flow. I suspect that may be something that can't be done due to the way the packets need to flow through the Trend code blobs? (dunno, just speculating).

Agreed, but what interests me is that Trend must be actively marking traffic on my router with Asus Adaptive QoS off and cake running as I still see this in traffic analyser:

I am sure something along these lines would be useful to add to cake. Some basic tagging would go a long way to allowing us to meaningfully use one of the diffserv settings (I think 4 is plenty).

It seems like something that is going to put more load on the CPU, so those with fast connections may want the ability to stick with a tagless besteffort setup, or maybe limit the tagging to uploads only.

I agree. It tends to be those with slower connections that benefit most from traffic prioritisation anyway so probably the right approach.

There is a lot there and well above my pay-grade.

Sadly mine too. I would love to ask @dave14305 if he would be willing to take a look and see if he can make any sense of it, but I'm very aware that he must be max'ed out with FlexQoS currently.

 

[rgnldo]

with my NAS

NAS optimization.
Both edges must be processed.
With an AX88U + QNAP
we have a Wifi 6, quad core processor and a NAS.
It is to obtain this performance.

In my case, AC86U with USB 3.0 external hard drive plugged in.

 

[Smokey613]

Thanks Adamm! I did a reformat on my usb drive, selected reformat jffs on next boot, recreated a swap file, setup entware again and now the installer ran without any issues and Cake is up and running! I guess I just needed to clear out the cobwebs.

 

[rgnldo]

New project: own system: OpenBSD router.
Start today

 

[ugandy]

Peformance in samba?
75 gb in 42 minutes with wifi?
Is it the effect of Cake-QOS?

View attachment 24489

cake is only applied to the eth0 interface, right?

why would it affect flows that are within the wifi interface(s) only ? that is, between wifi clients and/or router?

in any case, if you repeat the copy operation with cake off, do you get a faster transfer?

 

[Wade Coxon]

cake is only applied to the eth0 interface, right?

why would it affect flows that are within the wifi interface(s) only ? that is, between wifi clients and/or router?

in any case, if you repeat the copy operation with cake off, do you get a faster transfer?

Presumably the NAS is on the eth0 interface and the client is on wifi.

That's a bit of a gocha if that's what's happening! Shame though since the wifi to LAN traffic shouldn't require routing. That will need sorting!

I haven't noticed it myself as I don't have any high bandwidth clients connecting directly to that wifi point.