CakeQOS CakeQOS-Merlin

[ttgapers]

I have two routers, both connected to Comcast docsis 3.1, same modem - same router. One is a home connection 250Mbit/ 12Mbit and the other is a business line with 75/25. The home connection works perfect with "unlimited" set, but when using "unlimited" on the business line it really throws off my BufferBloats. Here is a test on the business line with the following settings:

besteffort docsis unlimited ack-filter dual-srchost

View attachment 24439

and here is the same without "unlimited":

besteffort docsis ack-filter dual-srchost

View attachment 24440
Just wanted to put that out there. You really have to check carefully to be sure what settings work best for you. Took me a couple of hours to fine tune my connections. Here are my final settings
At Home 250/12:
Down
docsis unlimited ack-filter dual-dsthost
Up
docsis unlimited ack-filter dual-srchost

At Office 75/25:
Down
docsis ack-filter dual-dsthost
Up
docsis ack-filter dual-srchost

Agreed. Testing with this today

But always want to re-emphasize the beauty of CAKE is KISS - little to no config/tweaks required. All extra features are for the outlier users who require more tweaking than the majority of users who will just work with besteffort and ATM parameters (based on connection type).

Happy Canada Day for the 1.0.2 release!

 

[heysoundude]

I did not, please log an issue on Github about it.

As you wish.
https://github.com/ttgapers/cakeqos-merlin/issues/53

 

[ttgapers]

As you wish.
https://github.com/ttgapers/cakeqos-merlin/issues/53

Thanks man. If/when one has an issue, I prefer it gets logged there for triaging. Note, it's not to be used as a "Helpdesk", that's what here is for. Once something is an issue affecting all aka "reproducible" then let's get it there.

I probably didn't see it as I updated from a different branch than master last night.

Thanks again!

 

[heysoundude]

If/when one has an issue, I prefer it gets logged there for triaging. Note, it's not to be used as a "Helpdesk", that's what here is for. Once something is an issue affecting all aka "reproducible" then let's get it there.

Agreed!
You heard it here first folks: log issues on GitHub; this is for support.


Sent from my iPhone using Tapatalk

 

[ttgapers]

Agreed!
You heard it here first folks: log issues on GitHub; this is for support.


Sent from my iPhone using Tapatalk

Yep I'm doing my best to log them as I see them here. It helps devs keep track, if even to park some good ideas to longer term releases. The new features in this release was a prime example if you check the issue, and see the back/forth convo related to it becoming a reality.

1. https://github.com/ttgapers/cakeqos-merlin/issues/46
2. https://github.com/ttgapers/cakeqos-merlin/issues/42

Cheers!

 

[XIII]

Finally found time to try out CakeQOS-Merlin, but I'm afraid I'm doing it wrong...

Before installation the DSLReports test reported my 250 Mbps down / 50 Mbps up like this:

• Down 200 megabit/s
• Up 20 megabit/s
• Overall B
• Bufferbloat B

After installation (download 238, upload 23, besteffort, docsis ack-filter) it reports this:

• Down 189 megabit/s
• Up 18 megabit/s
• Overall B
• Bufferbloat B

I was expecting A's...

What's wrong? (My expectations?)

PS: This is with number of streams set to 4 on the test site (if I don't set it 18 out of 24 streams fail and I score C's)

 

[XIII]

I seem to be able to increase the number streams to 12 (16 gives errors again; did not try in between) and then I get 200/8/A/A.

(and unlike the previous test this is with Suricata installed/enabled)

But in a second test it's 200/15/B/B... Why so much fluctuation?

 

[ugandy]

I seem to be able to increase the number streams to 12 (16 gives errors again; did not try in between) and then I get 200/8/A/A.

(and unlike the previous test this is with Suricata installed/enabled)

But in a second test it's 200/15/B/B... Why so much fluctuation?

not sure but i've seen that too.
i actually prefer to use spdmerlin/pingplotter to check my bufferbloat
maybe related to the dslreports servers load/location?

 

[Cam]

ideally, could cake packages get into entware? i'm not familiar with that process either.

Not possible as Entware is built on 2.6 or 3.2 kernel - I think someone will have to compile from Rmerlins repo/toolchain

 

[ttgapers]

Finally found time to try out CakeQOS-Merlin, but I'm afraid I'm doing it wrong...

Before installation the DSLReports test reported my 250 Mbps down / 50 Mbps up like this:

• Down 200 megabit/s
• Up 20 megabit/s
• Overall B
• Bufferbloat B

After installation (download 238, upload 23, besteffort, docsis ack-filter) it reports this:

• Down 189 megabit/s
• Up 18 megabit/s
• Overall B
• Bufferbloat B

I was expecting A's...

What's wrong? (My expectations?)

PS: This is with number of streams set to 4 on the test site (if I don't set it 18 out of 24 streams fail and I score C's)

Have a look at the tips here: https://github.com/ttgapers/cakeqos-merlin#tips

Use 90-95% of your total/contracted up/down speeds. Therefore if you are 100/10 up/down, set 95/9 for example.

 

[ttgapers]

Not possible as Entware is built on 2.6 or 3.2 kernel - I think someone will have to compile from Rmerlins repo/toolchain

@Cam would you know/be able to guide how to do this? It is last on the outstanding items to get the other 2 possible supported routers in the list.

Cheers.

 

[ugandy]

Have a look at the tips here: https://github.com/ttgapers/cakeqos-merlin#tips

Use 90-95% of your total/contracted up/down speeds. Therefore if you are 100/10 up/down, set 95/9 for example.

90-95% of measured max speeds (hopefully they don't change much), not contracted speeds.

 

[ugandy]

@Cam would you know/be able to guide how to do this? It is last on the outstanding items to get the other 2 possible supported routers in the list.

Cheers.

and keeping all up to date...

 

[ttgapers]

90-95% of measured max speeds, not contracted speeds.

This is where this always get dicey lol. I get my contracted speed, so tend to use it. Others like yourself, have wild fluctuations so measured max I best in that case.

 

[Cam]

@Cam would you know/be able to guide how to do this? It is last on the outstanding items to get the other 2 possible supported routers in the list.

Cheers.

Love to say yes, but haven't compiled a kernel module for some time so not sure I have the time to experiment without some tips from @Odkrys sorry.

 

[rgnldo]

Implement presets function
Important function for lay users, the function with predefined options with unlimited, rate DL / UP = 0, based on connectivity (DOCIS, VDSL etc)
Perhaps it is possible. Finding consensus with Cake-QOS users and the wan connectivity method.
Ideally, the band should be "0", on unlimited.

 

[brummygit]

Implement presets function
Important function for lay users, the function with predefined options with unlimited, rate DL / UP = 0, based on connectivity (DOCIS, VDSL etc)
Perhaps it is possible. Finding consensus with Cake-QOS users and the wan connectivity method.
Ideally, the band should be "0", on unlimited.

I think that the concept of adopting some defaults is a good idea

• unlimited bandwidth is the cake default anyway so we could adopt that - we must make it clear to people that this turns off the bandwidth shaping and therefore bufferbloat will not be controlled as well as setting fixed bandwidth values
• I am not convinced we should adopt a standard overhead as there will be many different connection types amongst the userbase - instead we could continue. as we do already, to not force it to be specified so we get the cake default of raw (none)
• I would agree that besteffort is a good default
• I find ack-filter useful so that gets my vote

Something else from the previous changes to allow parameters to be set independently on unload and download, I would find it useful to have 3 extra options sets:

1. Common (applies to upload and download)
2. Upload
3. Download

 

[rgnldo]

I think that the concept of adopting some defaults is a good idea

• unlimited bandwidth is the cake default anyway so we could adopt that - we must make it clear to people that this turns off the bandwidth shaping and therefore bufferbloat will not be controlled as well as setting fixed bandwidth values
• I am not convinced we should adopt a standard overhead as there will be many different connection types amongst the userbase - instead we could continue. as we do already, to not force it to be specified so we get the cake default of raw (none)
• I would agree that besteffort is a good default
• I find ack-filter useful so that gets my vote

Something else from the previous changes to allow parameters to be set independently on unload and download, I would find it useful to have 3 extra options sets:

1. Common (applies to upload and download)
2. Upload
3. Download

It will not be an adopted standard. Only the presets option. Manual configuration will continue.

 

[ttgapers]

It will not be an adopted standard. Only the presets option. Manual configuration will continue.

I am happy with the current preset option besteffort triple-isolate etc. The current additional flexibility/manual config provided I think is sufficient for the advanced users. We need defaults (which we currently have), where users can get it installed and going with little fuss and input (KISS). As such overheads shouldn't be a default, but something a user HAS to configure since everyone's internet connection is different. If along that path they want more advanced config options, they are available now at a more granular level.

 

[brummygit]

It will not be an adopted standard. Only the presets option. Manual configuration will continue.

Sorry I replied in a hurry, could have been clearer and used the word "standard" when I meant "default"

To be clear I think we can document some defaults which users can adopt, not force them on people using the script. Some people will still copy any suggestions into the fields without research which is why I think we need to be careful with any advice.