CakeQOS CakeQOS-Merlin

  • ATTENTION! You'll notice a Prefix dropdown when you create a thread. If your post applies to one of the topics listed, please use that Prefix for your post. When browsing the thread list you can use the Prefix to filter the view.
  • ATTENTION! As of November 1, 2020, you will not be able to reply to threads 6 months after the thread is opened. Threads will not be locked, so posts may still be edited by their authors.

Morris

Regular Contributor
I'd missed this originally, and tested today - it does indeed seem to still prioritise some traffic but on my download it's very minimal anyway. I added 'wash' to upload and it still assigned traffic to the Voice and Video tins. Thanks

No, no factory reset as it shouldn't be needed but I have been withdrawing consent from the privacy page after turning everything off.

I think I was doing something to cause it. Maybe it was because I was leaving my SSH session running the cake-qos script so I could easily check if it was running and look at my diffserv4 tin stats, but this was stopping it restarting cake when needed?


The video posted by @Morris states that cake can be reconfigured on the fly without any packet loss. I've certainly seen no ill effects.
I've also seen no ill effects making changes on the fly. I've also gone to using "besteffort" based on that video, set my speeds to the ones Verizon guarantees me as they provide headway and then beat up my like. 4 - 4K videos on my computer, as well as speed test, and a constant ping to 8.8.8.8. Also a VOIP call, and my son streaming video on his computer. I expected to run out of bandwidth on my 100-Mb symmetric link yet the only sign of a load was the slower than usual speed test. I'm done configuring, it just works.

Morris
 

heysoundude

Very Senior Member
Yes, this would be very helpful with recommendations for different common connection types and speeds. The install is supposed to be as easy as cake yet it is hardly that way due to the scattered documentation, contradictions made by different forum members as lots of assumptions are being made.

Morris
we should make up a logic tree, with some broad generalities that summarize the manpage, don't you think? That should get people in a proverbial zone, and give them some ideas on how to further tweak/customize a solution for their connection/network
 

heysoundude

Very Senior Member
I've also seen no ill effects making changes on the fly. I've also gone to using "besteffort" based on that video, set my speeds to the ones Verizon guarantees me as they provide headway and then beat up my like. 4 - 4K videos on my computer, as well as speed test, and a constant ping to 8.8.8.8. Also a VOIP call, and my son streaming video on his computer. I expected to run out of bandwidth on my 100-Mb symmetric link yet the only sign of a load was the slower than usual speed test. I'm done configuring, it just works.

Morris
YMMV, but I'm seeing good results with the rtt adjustment. not surprising results, but the network overall feels a hair more agile to me.
 

Morris

Regular Contributor
YMMV, but I'm seeing good results with the rtt adjustment. not surprising results, but the network overall feels a hair more agile to me.
The major results are when the link fills. I'm not shore if you would notice anything when the link hits 75% unless it's seeing traffic bursts to 100% as that's when cake gets involved. At least that's what I got out of the video.

Morris
 

Morris

Regular Contributor
we should make up a logic tree, with some broad generalities that summarize the manpage, don't you think? That should get people in a proverbial zone, and give them some ideas on how to further tweak/customize a solution for their connection/network
Engineers will get that yet there seem to be a fare amount of instruction followers (Script people) on the forums that need everything spelled out for them.

I suspect what is throwing some of us off is that we are coming from QOS environments where one had to constantly improve the classification and cake hides the classification internally.

Morris
 

Wade Coxon

Regular Contributor
I suspect what is throwing some of us off is that we are coming from QOS environments where one had to constantly improve the classification and cake hides the classification internally.
Not that our AsusWrt implementation is doing any meaningful classification yet. The really amazing thing is that Cake is still working brilliantly with all traffic being treated equally. Imagine how well it will work and properly classified traffic!
 

Morris

Regular Contributor
Not that our AsusWrt implementation is doing any meaningful classification yet. The really amazing thing is that Cake is still working brilliantly with all traffic being treated equally. Imagine how well it will work and properly classified traffic!
According to the video I posted, it works so well because it's not wasting CPU cycles classifying traffic the traditional way. It simply cares about streams that don't require the full pipe vs applications that fill the pipe as they are bulk transfers. Super simple and I've become a believer.

Morris
 

heysoundude

Very Senior Member
Engineers will get that yet there seem to be a fare amount of instruction followers (Script people) on the forums that need everything spelled out for them.

I suspect what is throwing some of us off is that we are coming from QOS environments where one had to constantly improve the classification and cake hides the classification internally.

Morris
yes (sadly) and yes - teaching digital communication theory overview would be beyond my interest other than to say "cake simply synchronises the bucket brigade - make sure you've got the buckets as full as possible without spillage, and keep them moving at the best pace, both upstream to and down from the various fires, and they will get put out efficiently". beyond that, move closer to a faster server (the fire hall) for lower latency in putting the fire out.
 

Morris

Regular Contributor
I was getting repeated kernel snap shots on RT-AC86U with Merlin 384.18 + Cake-QOS. Reported in this thread on the Merlin forum that is Trend Micro and to turn it off. No change

I then stopped cake-qos and the snaps stopped. I started googling cake-qos and found some places that talk about cake-qos as bandwidth limiting. This video sent me on the path to Asus Traditional QoS:

I set up traditional QoS up setting Queue Discipline to fq_codel. Limited bandwidth to 90% of my measured bandwidth and did not set WAN packet overhead as Verizon FIOS is effectively Ethernet. Next I ran the DSLReports Speed Test and got grades of A+ A+ A+.

After this I loaded my link with 4K video, voip, downloads, google video conference and even with this no stutter in video or voice and a DSLReports Speed Test and got grades of A+ A+ A+

I'm a happy camper...

Morris
 

Kingp1n

Very Senior Member
I was getting repeated kernel snap shots on RT-AC86U with Merlin 384.18 + Cake-QOS. Reported in this thread on the Merlin forum that is Trend Micro and to turn it off. No change

I then stopped cake-qos and the snaps stopped. I started googling cake-qos and found some places that talk about cake-qos as bandwidth limiting. This video sent me on the path to Asus Traditional QoS:

I set up traditional QoS up setting Queue Discipline to fq_codel. Limited bandwidth to 90% of my measured bandwidth and did not set WAN packet overhead as Verizon FIOS is effectively Ethernet. Next I ran the DSLReports Speed Test and got grades of A+ A+ A+.

After this I loaded my link with 4K video, voip, downloads, google video conference and even with this no stutter in video or voice and a DSLReports Speed Test and got grades of A+ A+ A+

I'm a happy camper...

Morris
Are you using default rules/settings inside traditional QoS?
 

Wade Coxon

Regular Contributor

Morris

Regular Contributor
Are you using default rules/settings inside traditional QoS?
I have only configured what I have described. When presented the page to allocate per application I closed it.

Morris
 

Morris

Regular Contributor
Did you reboot after withdrawing permission. Hard reboot with power-off?
The thing that is crashing is the Trend-Micro dcd service which should not be running at all when cake is running with Trend permission withdrawn.
As soon as I turned off the Trend Micro features the memory was returned. I did not reboot till after removing cake and continuing to see the snaps. I did not power down and no process should survive a restart. If one dose there is something seriously wrong with the design or there is use of uninitialized memory a serious software bug.

I plan to reintroduce TM in the afternoon as there have been no snaps since the reboot.

Morris
 

Yooshaw

Occasional Visitor
I was having problems with seeing live video feeds from 2 x Ring Cameras (connected via Wi-Fi). Whenever I went to see a live video feed, the video would not start. However, the video did upload to the Ring Servers, and I can go and see that footage. I narrowed it down to Cake. When I turned off Cake, the live video feed immediately appeared.

After some experimenting, I found that if I set "nonat", the issue was corrected, and I am able to see live video feed. I'm using besteffort, and everything else seems like it was before. Is there any adverse side affects from setting "nonat"?
 

brummygit

Senior Member
I was having problems with seeing live video feeds from 2 x Ring Cameras (connected via Wi-Fi). Whenever I went to see a live video feed, the video would not start. However, the video did upload to the Ring Servers, and I can go and see that footage. I narrowed it down to Cake. When I turned off Cake, the live video feed immediately appeared.

After some experimenting, I found that if I set "nonat", the issue was corrected, and I am able to see live video feed. I'm using besteffort, and everything else seems like it was before. Is there any adverse side affects from setting "nonat"?
This is very interesting, thanks for posting the question. I raised that I was having issues with my Ring Doorbell and Live View here https://www.snbforums.com/threads/release-cakeqos-merlin.64800/post-606715

I'm going to try your fix and see how it works for me as I'm in a double NAT environment anyway so I don't think the 'nat' parameter is helping in my use case

EDIT: Proving will take a while as Live View generally works for me although has periods of resulting in a black screen, particularly after a motion alert
 
Last edited:

Yooshaw

Occasional Visitor
This is very interesting, thanks for posting the question. I raised that I was having issues with my Ring Doorbell and Live View here https://www.snbforums.com/threads/release-cakeqos-merlin.64800/post-606715

I'm going to try your fix and see how it works for me as I'm in a double NAT environment anyway so I don't think the 'nat' parameter is helping in my use case
What's interesting is that it was my 2 x Flood Light Cams have this issue, but My Doorbell 2 was working fine. Now that it's working, I'm just hope that setting nonat is not causing besteffort to not function properly. I was looking over the Cake documentation, and seems that nonat is the default setting.
 

Wade Coxon

Regular Contributor
As soon as I turned off the Trend Micro features the memory was returned. I did not reboot till after removing cake and continuing to see the snaps. I did not power down and no process should survive a restart. If one dose there is something seriously wrong with the design or there is use of uninitialized memory a serious software bug.
I have encountered many instances (pre-cake) were I could not apply firmware updates, even after a soft reboot, and only a hardware power reboot would work. I suspect that the sift reboot doesn't sufficiently re-initialise some hardware components, which can carry problems forward until they are fully reset.

I suspect that had you persisted with cake, and merely rebooted the router, those errors would have not popped up again. I have definitely never seen any in my logs.
 

brummygit

Senior Member
What's interesting is that it was my 2 x Flood Light Cams have this issue, but My Doorbell 2 was working fine. Now that it's working, I'm just hope that setting nonat is not causing besteffort to not function properly. I was looking over the Cake documentation, and seems that nonat is the default setting.
I've been trawling the internet and I reached a few conclusions:
  1. Ring devices are generally poor and flawed - lots of people have issues without categoric fixes being known. There seem to be lots of placebos
  2. Moving to dedicated WiFi SSID helps - I currently have my Ring Pro Doorbell on it's own RT-AC68U in router mode with NAT turned off - I think the quiet network with minimal broadcast traffic helps as the Ring devices don't like busy, congested networks
  3. SIP plays a part - In NAT passthrough I have turned off the SIP helper which possibly helps
  4. The Live View problems are worse when motion is triggered so that the device is already uploading what it saw
  5. Ring support are not much help and won't give information about servers, ports and protocols out very readily
  6. Did I mention that Ring devices are generally poor and flawed?
Having said that I previously noted a definite correlation between Cake and my Live View challenges, although I currently have it working to a degree
 

Morris

Regular Contributor
I have encountered many instances (pre-cake) were I could not apply firmware updates, even after a soft reboot, and only a hardware power reboot would work. I suspect that the sift reboot doesn't sufficiently re-initialise some hardware components, which can carry problems forward until they are fully reset.

I suspect that had you persisted with cake, and merely rebooted the router, those errors would have not popped up again. I have definitely never seen any in my logs.
It dose not feel like a hardware issue as everything is working normal except for the log events. Of cause one never knows.

Now as far as cake goes, after dong a bunch of reading I'm wondering if the bandwidth limiting with fq_codel is cake is one and the same. Native Wort has it as a standard package. The wording "Traditional QOS" appears to be an Asus term. Historically (Traditionally) QOS was implemented by classifying and tagging. I posted the history of the issue I was having yet the use of bandwidth limiting is probably the more interesting thing I'm presenting. Possibly I should have done this as separate posts.

Morris
 

Mutzli

Very Senior Member
I've been trawling the internet and I reached a few conclusions:
  1. Ring devices are generally poor and flawed - lots of people have issues without categoric fixes being known. There seem to be lots of placebos
  2. Moving to dedicated WiFi SSID helps - I currently have my Ring Pro Doorbell on it's own RT-AC68U in router mode with NAT turned off - I think the quiet network with minimal broadcast traffic helps as the Ring devices don't like busy, congested networks
  3. SIP plays a part - In NAT passthrough I have turned off the SIP helper which possibly helps
  4. The Live View problems are worse when motion is triggered so that the device is already uploading what it saw
  5. Ring support are not much help and won't give information about servers, ports and protocols out very readily
  6. Did I mention that Ring devices are generally poor and flawed?
Having said that I previously noted a definite correlation between Cake and my Live View challenges, although I currently have it working to a degree
What are the problems you have with Ring cameras? I have a doorbell camera and 3 stick-up cameras and don't have any issues with cake or the router giving remote access to the cameras.
 

Similar threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top