What's new

CakeQOS CakeQOS-Merlin

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

after installed you can run with something like:
/jffs/addons/util/betterspeedtest.sh -t 60 -H netperf-west.bufferbloat.net -p 8.8.8.8

Thank you :)

It seems a cap on the download improves things. However adding docsis doesn't change things, even though I'm on cable. Nor did ack-filter. I'll have to keep testing. As of right now I've got some errors on some of the channels on my cable modem so maybe a refresh will help.

Thanks again!
 
Thank you :)

It seems a cap on the download improves things. However adding docsis doesn't change things, even though I'm on cable. Nor did ack-filter. I'll have to keep testing. As of right now I've got some errors on some of the channels on my cable modem so maybe a refresh will help.

Thanks again!
i found that ack-filter is mostly useful on the UL direction only, and only if your upload saturates easily.
 
This happened to me last night at midnight and my entire connection dropped - haven't had this happen before installing Cake. Also must the USB remain connected after installation?
 

Attachments

  • Screenshot_20200729-103627_Brave.jpg
    Screenshot_20200729-103627_Brave.jpg
    16.8 KB · Views: 152
Little curious about the settings. I added docsis both up and down due to seeing some PL in/out in a specific game. Also ran spMerlin and seen I have 63ms jitter? That's pretty high I imagine. But anyways I copied my settings to share cause I'm curious if anything is wrong. Download has wash. Upload has no-wash.

Code:
CakeQOS-Merlin: > Download Status:
qdisc cake 8008: dev ifb9eth0 root refcnt 2 bandwidth 45Mbit besteffort triple-isolate nat wash ingress no-ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

CakeQOS-Merlin: > Upload Status:
qdisc cake 8007: dev eth0 root refcnt 2 bandwidth 2700Kbit besteffort triple-isolate nat nowash no-ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

Anything look out of the ordinary? Also what is no-sce? Looked on that page and haven't found it.
 
Little curious about the settings. I added docsis both up and down due to seeing some PL in/out in a specific game. Also ran spMerlin and seen I have 63ms jitter? That's pretty high I imagine. But anyways I copied my settings to share cause I'm curious if anything is wrong. Download has wash. Upload has no-wash.

Code:
CakeQOS-Merlin: > Download Status:
qdisc cake 8008: dev ifb9eth0 root refcnt 2 bandwidth 45Mbit besteffort triple-isolate nat wash ingress no-ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

CakeQOS-Merlin: > Upload Status:
qdisc cake 8007: dev eth0 root refcnt 2 bandwidth 2700Kbit besteffort triple-isolate nat nowash no-ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

Anything look out of the ordinary? Also what is no-sce? Looked on that page and haven't found it.
I recommend you test/compare again with and without cake enabled. I bet you'll notice higher latency with cake disabled. Let us know...
 
With Cake-qos
Code:
Download: 41.78 Mbps
  Latency: (in msec, 60 pings, 0.00% packet loss)
      Min: 22.078
    10pct: 24.236
   Median: 28.451
      Avg: 28.670
    90pct: 32.848
      Max: 36.635
.............................................................
   Upload: 2.54 Mbps
  Latency: (in msec, 61 pings, 0.00% packet loss)
      Min: 25.095
    10pct: 25.353
   Median: 31.261
      Avg: 31.159
    90pct: 36.108

spMerlin with Cake for you
Code:
    Latency:    47.31 ms   (0.62 ms jitter)
   Download:    42.27 Mbps (data used: 47.4 MB)
     Upload:     2.56 Mbps (data used: 2.7 MB)
Packet Loss:     0.0%


With-out Cake-qos: no qdisc at all, Runner and Flow Cache enabled
Code:
Download: 48.89 Mbps
  Latency: (in msec, 60 pings, 0.00% packet loss)
      Min: 25.434
    10pct: 61.104
   Median: 75.683
      Avg: 73.911
    90pct: 83.710
      Max: 85.919
............................................................
   Upload: 2.94 Mbps
  Latency: (in msec, 61 pings, 0.00% packet loss)
      Min: 24.920
    10pct: 32.486
   Median: 48.108
      Avg: 49.318
    90pct: 61.128

spMerlin - no Cake for you
Code:
    Latency:    49.25 ms   (0.31 ms jitter)
   Download:    50.05 Mbps (data used: 71.9 MB)
     Upload:     3.21 Mbps (data used: 3.2 MB)
Packet Loss:     0.0%
 
Little curious about the settings. I added docsis both up and down due to seeing some PL in/out in a specific game. Also ran spMerlin and seen I have 63ms jitter? That's pretty high I imagine. But anyways I copied my settings to share cause I'm curious if anything is wrong. Download has wash. Upload has no-wash.

Code:
CakeQOS-Merlin: > Download Status:
qdisc cake 8008: dev ifb9eth0 root refcnt 2 bandwidth 45Mbit besteffort triple-isolate nat wash ingress no-ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

CakeQOS-Merlin: > Upload Status:
qdisc cake 8007: dev eth0 root refcnt 2 bandwidth 2700Kbit besteffort triple-isolate nat nowash no-ack-filter split-gso rtt 100.0ms noatm overhead 18 mpu 64 no-sce

Anything look out of the ordinary? Also what is no-sce? Looked on that page and haven't found it.
I don't see anything unusual in your settings, however I have recently had an issue myself which may be similar or may not.

I have a Ring Doorbell Pro in my home. These do seem to be quite fussy regarding wifi signal and other factors but I hit a real problem that the Live View function where you view the live video feed from either laptop or mobile phone would say it had connected but usually just stuck with a black screen, or occasionally connected with view poor quality. I did lots of troubleshooting but to keep the story short I discovered the following:
  • With cakeqos active my Ring Live View problem was there
  • If I stopped cakeqos Ring Live View worked fine
  • If I restarted cakeqos with bandwidth limits set to unlimited Ring Live View was fine
I haven't finished diagnosing my problem, but have been running a few days using FlexQoS (ie Asus Adaptive QoS) and so far I still don't see the issue. Once I have proven I can go a week without issues, I will swap back to cakeqos and test further.

I would be interested to know if setting both bandwidth values as 0 improve your problem - I know it will increase bufferbloat, but it would be interesting to hear your result as a diagnosis aid.

EDIT: I should add that Ring Live View gets its live feed via the internet so the traffic is leaving my network in the hop from doorbell to Ring cloud services, then straight back into my network again to the device running the Ring client.
 
I think I'm reading the results incorrectly. The jitter is in decimal point not 63ms, it's 0.63ms, I should have wrote 0.31-0.63. And the higher jitter number was from a further location so I did a bunch of wrong, oof
 
I don't see anything unusual in your settings, however I have recently had an issue myself which may be similar or may not.

I have a Ring Doorbell Pro in my home. These do seem to be quite fussy regarding wifi signal and other factors but I hit a real problem that the Live View function where you view the live video feed from either laptop or mobile phone would say it had connected but usually just stuck with a black screen, or occasionally connected with view poor quality. I did lots of troubleshooting but to keep the story short I discovered the following:
  • With cakeqos active my Ring Live View problem was there
  • If I stopped cakeqos Ring Live View worked fine
  • If I restarted cakeqos with bandwidth limits set to unlimited Ring Live View was fine
I haven't finished diagnosing my problem, but have been running a few days using FlexQoS (ie Asus Adaptive QoS) and so far I still don't see the issue. Once I have proven I can go a week without issues, I will swap back to cakeqos and test further.

I would be interested to know if setting both bandwidth values as 0 improve your problem - I know it will increase bufferbloat, but it would be interesting to hear your result as a diagnosis aid.

EDIT: I should add that Ring Live View gets its live feed via the internet so the traffic is leaving my network in the hop from doorbell to Ring cloud services, then straight back into my network again to the device running the Ring client.
Probably not related but my Nest Connect for my Nest Lock keeps losing connection when I enable DoT. Once disabled (DoT), I have to setup my Nest Connect again. I'm not sure if you're using DoT or not but something you may look at. I'm having no issues with cake-qos with any other device at this time.
 
Probably not related but my Nest Connect for my Nest Lock keeps losing connection when I enable DoT. Once disabled (DoT), I have to setup my Nest Connect again. I'm not sure if you're using DoT or not but something you may look at. I'm having no issues with cake-qos with any other device at this time.
Thanks, I don't use DoT but it's been like looking for a needle in a haystack to diagnose. I still don't think my problem is down to cake-qos but I am interested if @randomName can try my test just in case.
 
Are you curious about the Jitter?
I'm curious about the packet loss and jitter your experienced - and whether it's the traffic shaping which is switched off by setting the cake bandwidth limits to 0 (unlimited).

I have only had issues that I know of using the Ring Live View application, but it could be that we have an unexpected side effect of cake-qos which most applications cope with, but causes a few poor ones shows problems.
 
I'm curious about the packet loss and jitter your experienced - and whether it's the traffic shaping which is switched off by setting the cake bandwidth limits to 0 (unlimited).

I have only had issues that I know of using the Ring Live View application, but it could be that we have an unexpected side effect of cake-qos which most applications cope with, but causes a few poor ones shows problems.

It's hard to say cause the game I'm talking about where I see the PL most people see problems too. The game has been in Pre-Alpha for 4 years and has been abandoned by Epic. On the other hand I don't see PL with CS:GO. Contrary to the pre-alpha game observation with cake, I rarely seen it when I was running traditional QoS, and I mean rarely. As far as directly answering your question and with the game from Epic, when I first installed cake I didn't have the rates set correctly and I was seeing PL so Unlimited & 90-95% cap still returns PL up & down with that particular game. My ISP is 50/3 and I'm using 45/2.7


The jitter was in 0.31ms & 0.63ms The increase in jitter was due to a different location further away, and the jitter is in decimal point not a full 31ms or 63ms. That was my mistake. I've never seen a decimal point value with ms.
 
Last edited:
It's hard to say cause the game I'm talking about where I see the PL most people see problems too. The game has been in Pre-Alpha for 4 years and has been abandoned by Epic. On the other hand I don't see PL with CS:GO. Contrary to the pre-alpha game observation with cake, I rarely seen it when I was running traditional QoS, and I mean rarely. As far as directly answering your question and with the game from Epic, when I first installed cake I didn't have the rates set correctly and I was seeing PL so Unlimited & 90-95% cap still returns PL up & down. My ISP is 50/3 and I'm using 45/2.7


The jitter was in 0.31ms & 0.63ms The increase in jitter was due to a different location further away, and the jitter is in decimal point not a full 31ms or 63ms. That was my mistake. I've never seen a decimal point value with ms.
Great thanks, it helps with trying to find my own issue.
 
Great thanks, it helps with trying to find my own issue.

I might uninstall it and run traditional QoS to compare.. I'll report back when I feel I've got enough observational data/time with the game I get packet loss with.
 
People - for the volume of data travelling over the lengths of cable it does, your jitter numbers are about as good as you're going to get unless you go down the road of ntpMerlin with a satellite-based GPS time reference...with each machine on the server (and the server itself) referencing the same clock source...and even then you'll only get down into femtoseconds of jitter if you're lucky enough to own a unicorn and the blue moon is in the 7th house.
Packet Loss and latency indicate your connections aren't interfacing optimally- buffers aren't right, speeds are mismatched, priorities aren't correct...and these all rely on clocking (see above).
When you've gotten it as good as it can get with the gear you've got (or is available), talk to the people who code the software you're using about fault tolerance and improvement to those mechanisms. I'm talking the OS people and the people who write driver firmware as well as app coders, Those are the better places to chase down the dragons you're looking to slay
 
I have cake-qos installed and working fine. What I have noticed is that my IPTV service is terrible when my teenage son is downloading from the internet (usually from Steam) and the bandwidth is running at 100%- if where a way to make sure the IPTV device I have always has top priority and my son's PC is the lowest?
 
I have cake-qos installed and working fine. What I have noticed is that my IPTV service is terrible when my teenage son is downloading from the internet (usually from Steam) and the bandwidth is running at 100%- if where a way to make sure the IPTV device I have always has top priority and my son's PC is the lowest?
Not with cake unfortunately. The algorithm ensures fair bandwidth sharing across devices and streams with no priority for specific devices or types of traffic. This works very well except where your bandwidth is oversubscribed and you need specific devices to receive more than their fair share of it.

It's a job for Adaptive QoS (even better with FlexQoS or FreshJR) in these circumstances, or some people use traditional QoS.
 
It's a job for Adaptive QoS (even better with FlexQoS or FreshJR) in these circumstances, or some people use traditional QoS.
Cake can deal with tagged and classified data, however, this implementation for Merlin firmware doesn't include any classification (yet).

It would be great if someone could look into what is being done at OpenWRT and translate some of those tagging mechanisms over to here too. That way, we could actually properly use one of the diffserv modes rather than being limited to besteffort.

I posted earlier with some links to tagging efforts on other projects, but had no luck trying to reproduce them on my rt-ac86u. I don't personally have enough knowledge of the ins an outs of these routing/packet mangling protocols to do more than just try stuff to see what sticks. I am hoping someone who is actually clever can have a crack at it though!
 

Similar threads

Latest threads

Sign Up For SNBForums Daily Digest

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