CakeQOS CakeQoS-Merlin v2.0.0 alpha

[mikefrommanchester]

Hi - I have upgraded my RT-AC86U to 386.2_4 firmware with Cake-qos built in and activated it.

Is there any impact or do I need to do anything with regards to Cake-qos that I had installed via the Github onto USB using the following instructions:

GitHub - ttgapers/cakeqos-merlin: Custom add-on for supported Asus routers with Merlin firmware that introduces CAKE QoS as an alternative to Traditional/Adaptive/FreshJR implementations

Or does the new firmware with Cake-qos just overwrite this package...

I did stop it before I turned on Cake from the Merlin firmware and it seems to be working fine.

Just wanted to check I haven't got some conflict - that is going mess up something...

Thanks for any help

 

[Morris]

Hi - I have upgraded my RT-AC86U to 386.2_4 firmware with Cake-qos built in and activated it.

Is there any impact or do I need to do anything with regards to Cake-qos that I had installed via the Github onto USB using the following instructions:

GitHub - ttgapers/cakeqos-merlin: Custom add-on for supported Asus routers with Merlin firmware that introduces CAKE QoS as an alternative to Traditional/Adaptive/FreshJR implementations

Or does the new firmware with Cake-qos just overwrite this package...

I did stop it before I turned on Cake from the Merlin firmware and it seems to be working fine.

Just wanted to check I haven't got some conflict - that is going mess up something...

Thanks for any help

You should uninstall the old cake to:
- Recover the space
- have the latest code run

Morris

 

[Morris]

Wired, every device in my household that has to stream (Smart TV, AppleTV's, set top box, PS4 Pro, cameras, my sons laptop) in one way or another is connected by ethernet. The only exception is my own laptop, which I barely use for streaming content. I usually watch Netflix 4K with the app from the TV itself, which is connected with a 1Gb connection (shielded CAT6) to a switch and from there to my router. There is no other streaming device in use at the same time at the same Gigabit switch. Total distance: less dan 10m / 30ft. According to the speedtests run on the TV, even during peak hours, the connection itself / the available bandwidth shouldn't be the issue here.

I do the same. I suspect your issue may be ISP related. As the problem went away, they may have addressed the issue.

I hope it stays fixed,

Morris

 

[mikefrommanchester]

You should uninstall the old cake to:
- Recover the space
- have the latest code run

Morris

Thanks Morris - I have uninstalled and re-toggled Cake-qos from within Merlin firmware.

It seems to be working fine.

I have set the Wan Packet Overhead to PPPOE VDSL.

How can I see what version of the cake-qos package is running?

Is there a command from within the Terminal that I can call to list the installed packages and versions installed?

 

[Morris]

Thanks Morris - I have uninstalled and re-toggled Cake-qos from within Merlin firmware.

It seems to be working fine.

I have set the Wan Packet Overhead to PPPOE VDSL.

How can I see what version of the cake-qos package is running?

Is there a command from within the Terminal that I can call to list the installed packages and versions installed?

Any Time Mike,

My expertise is network engineering so I'll defer this question for one of our Linux gurus to answer.

Morris

 

[randomName]

Is there a command line to enable the CPU to run and full throttle, and one to set it back to default?

 

[Hazel]

Is there a command line to enable the CPU to run and full throttle, and one to set it back to default?

As far as I’m aware the CPU will be automatically utilized as needed, this is the default setting and I don’t know any commands to change that behavior on our routers, like changing scaling governors on other Linux systems. As I have read earlier in this thread, Cake is single-threaded and will use a single core to the max of it’s capabilities as required. Someone with more knowledge of CakeQoS, please correct me if I’m wrong.

 

[dave14305]

How can I see what version of the cake-qos package is running?

If you mean the Addon script, you can run cake-qos status and see the version number in the header.

If you mean the actual cake module in the firmware, there is no version associated with it. It comes from the same source code as the old cake though, so there is no difference other than how it was compiled.

This thread is for the CakeQoS-Merlin 2.0 Addon script and web interface. If you haven’t installed it, then you’re simply using the stock Cake implementation from Merlin (which is good enough for most people). Cake is Cake. CakeQOS generally refers to the Addon.

 

[dave14305]

To keep things moving along, there is an update to the alpha branch that will push everyone to the develop branch. In the develop branch, there are a couple minor changes to the script:

• Bring back a limited CLI menu system for installing, uninstalling, updating, status and debug.
• Automatically enable CAKE in the firmware if not yet enabled during install. This will also handle any 1.0.8 users being automatically upgraded to the built-in CAKE if they check for updates on 386.2 or greater.

No relevant changes to the behavior of CAKE itself, so don't be in a rush to update, but please do update at your leisure and let us know if it goes smoothly.

EDIT: You may have to update twice to get the latest code from develop.

 

[Mutzli]

'If you are shaping inbound, and cannot trust the diffserv
markings (as is the case for Comcast Cable, among others), it is
best to use a single queue "besteffort" mode with wash.'

Is this still true because this is what I have have enable on my download. And I am a Comcast customer (Non Cake Speeds - 89/18). Thanks.View attachment 33717

Try to enable Filter Duplicate TCP ACKs for uploads only. I'm also on Comcast and it improved my upload results.

 

[Mutzli]

To keep things moving along, there is an update to the alpha branch that will push everyone to the develop branch. In the develop branch, there are a couple minor changes to the script:

• Bring back a limited CLI menu system for installing, uninstalling, updating, status and debug.
• Automatically enable CAKE in the firmware if not yet enabled during install. This will also handle any 1.0.8 users being automatically upgraded to the built-in CAKE if they check for updates on 386.2 or greater.

No relevant changes to the behavior of CAKE itself, so don't be in a rush to update, but please do update at your leisure and let us know if it goes smoothly.

EDIT: You may have to update twice to get the latest code from develop.

Updated twice and everything works as expected. I really like that you brought back the CLI menu again.

 

[Hazel]

Following up on my previous post, I have made a side by side comparison with Filtering TCP ACKs on uploads enabled and disabled, prior to updating to the latest version.

Amount of uploaded data is nearly equal, there's a 2Gb difference in downloaded data. The right half screenshot was taken first (Filter duplicate TCP ACKS = No), the other half (Filter duplicate TCP ACKs = Yes) last.

I haven't experienced any lag on Netflix 4K/UHD any more, which could be caused by a change of settings (no more filtering duplicates), a temporary issue with my ISP (although speed tests are very stable), load balancing issues with Netflix, or some other factors.

Could someone with more knowledge check my screenshots and let me know if I have any benefit by filtering duplicate ACKs or should I just leave this setting disabled? I have 0 dropped ACK packets when it's disabled but I don't know who to interpret that. Is that a good thing or not?

Full res (readable) version available here:

https://imgur.com/nFVEKmM

Looking forward to learn something new and your opinions.

Thanks in advance.

 

[Wistuplu]

let me know if I have any benefit by filtering duplicate ACKs or should I just leave this setting disabled? I have 0 dropped ACK packets when it's disabled but I don't know who to interpret that. Is that a good thing or not?

Here are my 2 cts.
When you download stuff, you have to send back ACKs. If your link is very asymetrical (like 50 to 1, i.e. 50 times more throughput on the download), you might end up with throttling on the download speed/throughput because the link is not able to send all the expected ACKs back on the upload (on top of possibly other ongoing activities on the upload like videoconferencing) . "Filter duplicate ACKs" tries to improve things in those scenarios. Except that, as Dave Taht explained in the reference I provided, this applies to very asymetrical links. Not to a 10:1 link as yours. If I remember correctly they made an analysis in a serious paper with the 1:50 example. And Taht referred to 1:20 or 1:15 as the limit where testing "Filter Duplicate ACKs" would make sense.
My suggestion is to go with what you feel/experience works best on your network, and not pay more attention to statistics. I know I gave them too much attention ;-)
I'm not an expert in traffic control or Cake, just a guy with some background and some desire to understand. I might be completely misled ;-)
Take care
W.

 

[Morris]

Here are my 2 cts.
When you download stuff, you have to send back ACKs. If your link is very asymetrical (like 50 to 1, i.e. 50 times more throughput on the download), you might end up with throttling on the download speed/throughput because the link is not able to send all the expected ACKs back on the upload (on top of possibly other ongoing activities on the upload like videoconferencing) . "Filter duplicate ACKs" tries to improve things in those scenarios. Except that, as Dave Taht explained in the reference I provided, this applies to very asymetrical links. Not to a 10:1 link as yours. If I remember correctly they made an analysis in a serious paper with the 1:50 example. And Taht referred to 1:20 or 1:15 as the limit where testing "Filter Duplicate ACKs" would make sense.
My suggestion is to go with what you feel/experience works best on your network, and not pay more attention to statistics. I know I gave them too much attention ;-)
I'm not an expert in traffic control or Cake, just a guy with some background and some desire to understand. I might be completely misled ;-)
Take care
W.

The reason for intermediate ACK is to avoid the delays caused when the transmitting host has emptied it's buffer and is ready to send yet can not as the previous send has not been acknowledged. The TCP/IP Rotate window's purpose is to allow data in flight without having to wait for the packet to be acknowledged to send the next packet. If transmitting host, receiving host and all routers in the path and links in the path are not causing delays, then filtering ACKs may work. As the state of all of these systems are constantly changing and any one of them could result in delays this can lead to the transmitting host not being able to send as it's waiting on the one ACK that will come at the end of the entire rotate window. This can be seen as spinning wheels in video streams, choppy audio, and lag in a game.

If you uplink in a non symmetric connection dose not have the bandwidth to carry a few ACK packets, the fix is to provision more bandwidth. By filtering ACKs you are creating a new problem and it will be intermittent.

Morris

 

[randomName]

What's the best "flow isolation" for gaming when only 1 device on the router is being used? Fairness doesn't seem to apply in a situation like this, or am I wrong? Would it be Flowblind, if so?

 

[Robe]

Hi Folks. I am new to CakeQoS and was wondering how to set the up/down parameters for an ISP with significant sag during peak times i.e. 25->5 Mbps Down and 5->2 Mbps Up.

If I choke down the Cake settings to 12 Mbps Up and 3.5 Mbps Down, I can get excellent (A+) to OK (C) bufferbloat scores, but this is highly variable, severely limiting my top end, and does not work well at lower speeds. Is there a better way to manage these settings? I did see an earlier post, a page back, that referred to variable settings to address sag, but does this apply to this integrated Cake version and, if so, how would I go about setting this up to test it?

I am on a fixed wireless system using a radio on the roof, for ISP tower communications, which plugs directly into the WAN port on my router. It's not really Cable or DSL, so what setting would I use for this set up, or does it matter if the settings are correct?

I am using an AX88u router and AC68u node both with the latest Merlin firmware.

Thank you so much for your help and all the great work that is done behind the scenes by Merlin et al!

 

[pdc]

Hi Folks. I am new to CakeQoS and was wondering how to set the up/down parameters for an ISP with significant sag during peak times i.e. 25->5 Mbps Down and 5->2 Mbps Up.

If I choke down the Cake settings to 12 Mbps Up and 3.5 Mbps Down, I can get excellent (A+) to OK (C) bufferbloat scores, but this is highly variable, severely limiting my top end, and does not work well at lower speeds. Is there a better way to manage these settings? I did see an earlier post, a page back, that referred to variable settings to address sag, but does this apply to this integrated Cake version and, if so, how would I go about setting this up to test it?

I am on a fixed wireless system using a radio on the roof, for ISP tower communications, which plugs directly into the WAN port on my router. It's not really Cable or DSL, so what setting would I use for this set up, or does it matter if the settings are correct?

I am using an AX88u router and AC68u node both with the latest Merlin firmware.

Thank you so much for your help and all the great work that is done behind the scenes by Merlin et al!

I don't have direct experience with your situation, but I did bookmark an interesting discussion on how to handle variable bandwidth.

Start reading from: https://www.snbforums.com/threads/cakeqos-merlin.64800/post-597689

The idea is to try either unlimited, which will not take link bandwidth into account, or autorate-ingress which adjusts over time based on current bandwidth usage. The tradeoff seems to be that autorate-ingress adjusts more slowly to bandwidth demand, but controls bufferbloat/latency better.

See https://www.man7.org/linux/man-pages/man8/tc-cake.8.html for explanation of the options.

Give them a try and see what works for you...

As for your connection, per the man page you could start with conservative, which "Compensates for more overhead than is likely to occur on any widely-deployed link technology."

 

[Morris]

Hi Folks. I am new to CakeQoS and was wondering how to set the up/down parameters for an ISP with significant sag during peak times i.e. 25->5 Mbps Down and 5->2 Mbps Up.

If I choke down the Cake settings to 12 Mbps Up and 3.5 Mbps Down, I can get excellent (A+) to OK (C) bufferbloat scores, but this is highly variable, severely limiting my top end, and does not work well at lower speeds. Is there a better way to manage these settings? I did see an earlier post, a page back, that referred to variable settings to address sag, but does this apply to this integrated Cake version and, if so, how would I go about setting this up to test it?

I am on a fixed wireless system using a radio on the roof, for ISP tower communications, which plugs directly into the WAN port on my router. It's not really Cable or DSL, so what setting would I use for this set up, or does it matter if the settings are correct?

I am using an AX88u router and AC68u node both with the latest Merlin firmware.

Thank you so much for your help and all the great work that is done behind the scenes by Merlin et al!

You need to get everyone in town to complain to the ISP. Routers can't fix bad ISPs.

Good luck,

Morris

 

[joe scian]

Hi guys - Hotfix showing as available in GUI. When applied it still shows hotfix available.

 

[dave14305]

Hi guys - Hotfix showing as available in GUI. When applied it still shows hotfix available.

Update again. It probably has to do with promoting everyone from alpha to develop branch. This was mentioned back on May 8. If it’s still there, let us know.