CakeQOS CakeQOS-Merlin

[Mutzli]

I am still doubtful the effectiveness of Cake. According to https://www.bufferbloat.net/projects/codel/wiki/CakeTechnical/ the article state for differserv*:
"Most traffic falls into the Best Effort class. VoIP, NTP and gaming traffic should be directed to the Voice class, BitTorrent should be directed to the Bulk class, and the Video class is available for any bulk traffic that requires elevated priority."

I have only had seen a few bytes here and there going to Voice and 99.9% has all been going through Best Effort. While we have kids constantly on the Xbox gaming+chatting in game, in-laws streaming from my Plex Server(they are off-site) and gazillion bittorrent traffic. So far, it looks great for controlling bufferbloating, but it hasn't don't much traffic shaping.

I think FlexQoS would be a better solution for you if you want to have more control over the individual traffic shaping. Maybe sometimes in the near future we can have both, FlexQoS and Cake to control buffer bloat running at the same time.

 

[Wade Coxon]

do you have a /jffs/scripts/nat-start
with entry below:
/jffs/addons/cake-qos/cake-qos start & # CakeQOS-Merlin

Just checked and I have the same problem.

I do have a nat-start, but it is empty.

 

[brummygit]

I am still doubtful the effectiveness of Cake. According to https://www.bufferbloat.net/projects/codel/wiki/CakeTechnical/ the article state for differserv*:
"Most traffic falls into the Best Effort class. VoIP, NTP and gaming traffic should be directed to the Voice class, BitTorrent should be directed to the Bulk class, and the Video class is available for any bulk traffic that requires elevated priority."

I have only had seen a few bytes here and there going to Voice and 99.9% has all been going through Best Effort. While we have kids constantly on the Xbox gaming+chatting in game, in-laws streaming from my Plex Server(they are off-site) and gazillion bittorrent traffic. So far, it looks great for controlling bufferbloating, but it hasn't don't much traffic shaping.

I posted on the FlexQoS thread earlier today and was holding back here as focus is understandably on the v1.0.0 release, however here's pretty much what I posted there:

I'm one of those testing Cake, but whilst it's amazing at managing the queues, I still think there could be a future where Cake is configured with 4 or 8 "tins" using the traffic marking from Asus / Trend adaptive QoS and the FlexQoS script. My reasoning is that cake shares everything fairly, whereas I had issues yesterday with multiple people viewing Premier League football being interrupted due to my bandwidth being way over subscribed, but cake trying to share it fairly. What I needed was for the low priority traffic (a large download someone else was running) to back off and only get a much reduced share of the bandwidth. In other words as you say, traffic shaping.

Cake has "diffserv" which splits the traffic into up to 8 "tins" (essentially priority queues) but it's built in identification seems to put almost all of my traffic into it's default tin. I'm hoping that there's a way to improve this situation and shape the traffic.

Importantly, whilst testing cake I haven't withdrawn my consent from Trend Micro (contrary to advice on the thread) as I wanted to keep using other features including the bandwidth monitor and app analysis. I'm not bothered about memory use etc as long as my router runs fine. I have to say that cake and Trend are living happily together.

The icing on the cake would be a collaboration between the 2 projects to get the cake benefits enhanced by traffic prioritisation and shaping. I understand that focus will currently be on polishing Cake-QoS and FlexQoS both of which are developing quickly, but wanted to put my thoughts out there for future consideration.

 

[Stephen Harrington]

do you have a /jffs/scripts/nat-start
with entry below:
/jffs/addons/cake-qos/cake-qos start & # CakeQOS-Merlin

I do NOW, 3rd time lucky uninstall/reinstall seems to have done it, some unexplained weirdness going on!

Edit: Is it possible that the adding of the entry into nat-start is a bit “intermittent” given @Wade Coxon has the same issue?

 

[mbze430]

I think FlexQoS would be a better solution for you if you want to have more control over the individual traffic shaping. Maybe sometimes in the near future we can have both, FlexQoS and Cake to control buffer bloat running at the same time.

I believe the current problem is either our current Cake implementation is missing proper tagging or it is tagging everything and tossing in the BE tin when Diffserv* is enabled

I am not home to sniff the packets if DSCP is being tagged... but anyone that have the time here is an article from Wireshark:
https://blog.wireshark.org/2009/09/filtering-dscp/
CISCO's break down on DSPC bits
https://www.cisco.com/c/en/us/suppo...-qos/qos-packet-marking/10103-dscpvalues.html

 

[Wade Coxon]

I do NOW, 3rd time lucky uninstall/reinstall seems to have done it, some unexplained weirdness going on!

Edit: Is it possible that the adding of the entry into nat-start is a bit “intermittent” given @Wade Coxon has the same issue?

An uninstall and reinstall worked for me.

It may have been linked to my initial trouble with installing it due to my out of date package list.

 

[ugandy]

uninstall old, and install new, no errors.

had to manually clean services-stop

there is a typo:
"Select Queue Prioity Type:"

suggestions:
I think the "extraOptions" prompt, could offer some defaults like ack-filter, etc.
also maybe a separate choice for connection type: docsis, etc


the new version also has a watchdog, right?
after uninstall old and install new,

i have:

cromo@RT-AX88U-8158:/tmp/home/root# cru l|grep cake
*/60 * * * * /jffs/addons/cake-qos/cake-qos checkrun #CakeQOS-Merlin#


but nat-start does not have entry for cake!

 

[ugandy]

...could we get more detailed status information added to the menus as extra options. I'm thinking the

tc qdisc
tc qdisc show | grep root
tc -s qdisc show dev eth0 # for upload
tc -s qdisc show dev ifb9eth0 # for download

Commands from the GitHub page

this would be useful! maybe even parsed and presented in some simplified qualitative metric: green, yellow, red, kind of a thing? some webUI graph!

 

[ugandy]

before cake, I'd normally get a C or a D for bufferbloat
i still have the problem that comcast upload varies anywhere from 20Mbps to 5Mbps during the workday, but that's another story

 

[Wade Coxon]

Ok, I think I found the problem with nat-start

If you install the script, nat-start is populated correctly.

However, if you then adjust the cake settings via option 4 of the script menu, once you save the settings, the cake-qos entry is removed from nat-start start.

oops!!

 

[Torson]

Ok, I think I found the problem with nat-start

If you install the script, nat-start is populated correctly.

However, if you then adjust the cake settings via option 4 of the script menu, once you save the settings, the cake-qos entry is removed from nat-start start.

oops!!

I made a couple of setting changes to it post-install. The nat-start statement is still there, also the cru line is there. It works!
Edit - the bufferbloat result is spectacular

 

[maghuro]

Ok, I think I found the problem with nat-start

If you install the script, nat-start is populated correctly.

However, if you then adjust the cake settings via option 4 of the script menu, once you save the settings, the cake-qos entry is removed from nat-start start.

oops!!

Hot fixed thanks!

 

[ugandy]

did i mess up, or was there a change in the latest cake installer regarding HW accel?
after a fresh install i see that now i have runner off but flow cache on.
cake now reaches 330Mbps on my ax88

 

[Adamm]

did i mess up, or was there a change in the latest cake installer regarding HW accel?
after a fresh install i see that now i have runner off but flow cache on.
cake now reaches 330Mbps on my ax88

I fixed it so the nvram values for runner (which shows on the WebUI) sync with us manually disabling the feature. Flow cache is also disabled but I haven't made a similar change just yet until I do some further investigation.

 

[ugandy]

I fixed it so the nvram values for runner (which shows on the WebUI) sync with us manually disabling the feature. Flow cache is also disabled but I haven't made a similar change just yet until I do some further investigation.

thanks. still puzzled with my speed increase to 330 compared with yesterday's 280. maybe some side change on my setup that i didn't control properly.

 

[skeal]

had to manually clean services-stop

I have a cake-qos entry in services-stop should it be there?

 

[ugandy]

I have a cake-qos entry in services-stop should it be there?

yes. i was just removing the entry from the old version version before installing new version

 

[skeal]

Everytime I reboot nat-start clears itself and Cake won't start on a reboot.

 

[skeal]

Every reboot this entry in nat-start is removed.

/jffs/addons/cake-qos/cake-qos start & # CakeQOS-Merlin

 

[ugandy]

Everytime I reboot nat-start clears itself and Cake won't start on a reboot.

did you try cake-qos update/install before reboot?
there was an updated version on gitbub