CakeQOS CakeQOS-Merlin

[ttgapers]

New release v0.0.6
- Implemented watchdog (it'll be installed automatically when starting script)
- No need anymore to do install nor enable. When doing cake-qos start XXMBit XXMBit "param" the script will auto do the necessary things
- Readme updated with new instructions
- Bugfixes

Just remember that this is a WIP, so it MAY have bugs. Please test it and report it on github issues page. Thanks

Thanks...thread updated!

 

[Audionut]

tried autorate-ingress
it was applied to both UL and DL queues.
the rates were constantly changing. sometimes the UL limit was set to high and the DL limit to low.
not using for now.

Agreed. I wanted to try this myself and found the same as you. bufferbloat was horrible.

 

[Wade Coxon]

Apologies for bumping my own post, but I think it got lost at the end of page 12 during a flurry of other update activity.
https://www.snbforums.com/threads/release-cakeqos-merlin.64800/page-12#post-596028

In short, can cake handle enabling itself on upload without applying itself on the download queue. Handy to have for people whose download rates exceed 300Mbs, but still want upload queue management.

Suggest that the script interprets a 0Mbps rate as "disable cake on that queue".

 

[jackiechun]

Apologies for bumping my own post, but I think it got lost at the end of page 12 during a flurry of other update activity.
https://www.snbforums.com/threads/release-cakeqos-merlin.64800/page-12#post-596028

In short, can cake handle enabling itself on upload without applying itself on the download queue. Handy to have for people whose download rates exceed 300Mbs, but still want upload queue management.

Suggest that the script interprets a 0Mbps rate as "disable cake on that queue".

There may be a chance that cake has this built-in, it's worth a try I know at least for OpenWRT SQM if you leave one of the speeds as 0, it won't apply.

Try something like:

cake-qos start 0Mbit 30Mbit

 

[ugandy]

There may be a chance that cake has this built-in, it's worth a try I know at least for OpenWRT SQM if you leave one of the speeds as 0, it won't apply.

Try something like:

cake-qos start 0Mbit 30Mbit

it removed the DL limit:

CakeQOS-Merlin - v0.0.6: > Download Status:
qdisc cake 800a: dev ifb9eth0 root refcnt 2 bandwidth unlimited diffserv4 triple-isolate nat wash ingress ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce



but DL is still ~330Mbps

 

[Wade Coxon]

it removed the DL limit:

CakeQOS-Merlin - v0.0.6: > Download Status:
qdisc cake 800a: dev ifb9eth0 root refcnt 2 bandwidth unlimited diffserv4 triple-isolate nat wash ingress ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

but DL is still ~330Mbps

Yes, thanks, that's what I am talking about. It needs to remove cake from the interface entirely to remove the bottleneck.

 

[ugandy]

@L&LD did you get 400Mbps with cake on an AX88?

 

[L&LD]

@ugandy, closer to 5ooMbps, and that was with the initial release too.

https://www.snbforums.com/threads/release-cakeqos-merlin.64800/#post-595227

 

[SomeWhereOverTheRainBow]

@ugandy, closer to 5ooMbps, and that was with the initial release too.

https://www.snbforums.com/threads/release-cakeqos-merlin.64800/#post-595227

what options are you using (*example what did your enable command look like?) and did you disable adaptive qos per recommendations?

 

[L&LD]

@SomeWhereOverTheRainBow I'm not using it anymore. Please see the link I posted for more details.

 

[SomeWhereOverTheRainBow]

@SomeWhereOverTheRainBow I'm not using it anymore. Please see the link I posted for more details.

so your method was that you only used "docsis ack-filter" parameters? What made you decide to skip the besteffort parameter?

 

[jackiechun]

so your method was that you only used "docsis ack-filter" parameters? What made you decide to skip the besteffort parameter?

Oh in the early builds besteffort was forced upon everyone

 

[Wade Coxon]

Interesting. If the ax88u is faster, that means that cake must be using additional threads somewhat effectively then? (Since the individual cores and clock speeds are the same as the ac86u)

 

[SomeWhereOverTheRainBow]

Oh in the early builds besteffort was forced upon everyone

I am tempted to test diffserv4 and diffserv8

 

[Wade Coxon]

It looks like it still defaults to besteffort if none is specified.

Any reason to use that as default rather than use the cake diffserv3 default?

 

[jackiechun]

It looks like it still defaults to besteffort if none is specified.

Any reason to use that as default rather than use the cake diffserv3 default?

From: https://www.bufferbloat.net/projects/codel/wiki/Cake/

"We have generally found that most diffserv inbound priorities are wrong, so we tend to specify besteffort here"

Also apparently lighter on CPU usage (as it's just 1 "tin" vs 3, 4 or 8), but I'd be interested in seeing test comparisons between them.

 

[SomeWhereOverTheRainBow]

It looks like it still defaults to besteffort if none is specified.

Any reason to use that as default rather than use the cake diffserv3 default?

with best effort only selected
Besteffort effectively ignores DSCP values, every packet is as important as the other and so just flow & host fairness come into play.

The diffserv modes split DSCP into an increasing number of tins.

 

[SomeWhereOverTheRainBow]

By default CAKE uses a 3 tin classification mode called 'diffserv3'. This does not factor in the video class properly.

 

[ttgapers]

From: https://www.bufferbloat.net/projects/codel/wiki/Cake/

"We have generally found that most diffserv inbound priorities are wrong, so we tend to specify besteffort here"

Also apparently lighter on CPU usage (as it's just 1 "tin" vs 3, 4 or 8), but I'd be interested in seeing test comparisons between them.

On my todo list, I tried with 8 and switched back to besteffort

 

[SomeWhereOverTheRainBow]

On my todo list, I tried with 8 and switched back to besteffort

4 is probably the best situational one to use if you are not using best effort.