FlexQoS FlexQoS 1.2.4 - Flexible QoS Enhancement Script for Adaptive QoS

[jack ryan]

running 386.2 alpha on RT-ac5300 with flexqos 1.2.2 dev, it gives me worst jitters in upload.
jitter goes up to 1000 ms.
setting as below.

 

[Treadler]

My take on the 1.2.2 dev, bufferbloat (that I didn’t have any great problem with using the stable branch) yes, is improved.

 

[fallinnn]

Could someone answer something I've always wondered:
When manually setting internetspeed in QoS on my 500/500mbps net. Should I set the speed I get when close to router / wired on a laptop, or what I actually get from my gaming pc (wifi)? If I only get around 400/400 on my gaming pc, do I actually get the benefits of having QoS when the QoS is set to 500/500? Also: Any specific settings I should pick for FlexQoS to work smoothly on my higher speed net compared to default?

 

[el pescador]

Hi Dave. Thos instructions the other day regarding changing bursts etc...I couldnt get the figures to stick. Any chance of a little help? I know i must have done something wrong.

 

[dave14305]

running 386.2 alpha on RT-ac5300 with flexqos 1.2.2 dev, it gives me worst jitters in upload.
jitter goes up to 1000 ms.
setting as below.

View attachment 31332

Is it better on stable version 1.2.1?

 

[dave14305]

Could someone answer something I've always wondered:
When manually setting internetspeed in QoS on my 500/500mbps net. Should I set the speed I get when close to router / wired on a laptop, or what I actually get from my gaming pc (wifi)? If I only get around 400/400 on my gaming pc, do I actually get the benefits of having QoS when the QoS is set to 500/500? Also: Any specific settings I should pick for FlexQoS to work smoothly on my higher speed net compared to default?

My opinion is you set it based on the internet speeds you achieve while wired to the router, since QoS will be shaping traffic from all clients on the LAN. WiFi introduces its own latency into the equation, for sure, but we can’t really solve that on our own. There’s no harm in setting QoS based on 400/400, but if your client WiFi is the bottleneck, I don’t believe the lower setting will counteract that. You’ll just be throttling other devices that could have made use of the extra bandwidth.

 

[dave14305]

Hi Dave. Thos instructions the other day regarding changing bursts etc...I couldnt get the figures to stick. Any chance of a little help? I know i must have done something wrong.

Correct, they weren’t meant to stick. I was just suggesting where to find the tc commands FlexQoS generates. Then you can copy them to base your experiments on by running them manually over SSH after changing the values you wanted to change.

 

[el pescador]

Correct, they weren’t meant to stick. I was just suggesting where to find the tc commands FlexQoS generates. Then you can copy them to base your experiments on by running them manually over SSH after changing the values you wanted to change.

Ok thanks.
So for

class change dev br0 parent 1:1 classid 1:10 htb overhead 4 linklayer ethernet prio 0 rate 138240Kbit ceil 921600Kbit burst 115200b cburst 1152000b quantum 1728000

how would i run this in ssh?

 

[dave14305]

Ok thanks.
So for

class change dev br0 parent 1:1 classid 1:10 htb overhead 4 linklayer ethernet prio 0 rate 138240Kbit ceil 921600Kbit burst 115200b cburst 1152000b quantum 1728000

how would i run this in ssh?

tc class change dev br0 parent 1:1 classid 1:10 htb overhead 4 linklayer ethernet prio 0 rate 138240Kbit ceil 921600Kbit burst 115200b cburst 1152000b quantum 1728000

 

[el pescador]

tc class change dev br0 parent 1:1 classid 1:10 htb overhead 4 linklayer ethernet prio 0 rate 138240Kbit ceil 921600Kbit burst 115200b cburst 1152000b quantum 1728000

Thanks Dave.

 

[el pescador]

Dave. When i check the values in debug mode they havent changed.
I then try to run the grep again and no file exists.

 

[dave14305]

Dave. When i check the values in debug mode they havent changed.
I then try to run the grep again and no file exists.

Correct, this won't work with the script. All you're doing is manually changing the classes to see how they behave for you outside of the script. If you run flexqos start or debug, the tmp file will be regenerated (and after a debug it is deleted since it wasn't actually used to change anything).

 

[BoostOver]

Correct..

Currently traditional QoS is the best choice for me with multiplayer (I want to remind you that I have a 19/1 connection).
I am available to help you in this regard with your FlexQos
Thanks Dave

 

[dave14305]

Currently traditional QoS is the best choice for me with multiplayer (I want to remind you that I have a 19/1 connection).
I am available to help you in this regard with your FlexQos
Thanks Dave

Please post or PM me the contents of these commands when running Traditional QoS. You might need to upload it to a site like pastebin.com if it's too much data to post.

cat /tmp/qos
tc -s -d qdisc ls
tc -s -d class show dev br0 parent 2:
tc -s -d class show dev ppp0

In general, I think I need to port over more of the SQM functions from OpenWRT since there is a long history there dealing with low-bandwidth connections. You probably need the similar limitations on cburst, and perhaps increased values for fq_codel target and interval based on only 1 Mbit upload. But only if your qdisc stats show a lot of drops on ppp0.

Mathematically, at 849 kbps upload, it would take 14 ms to send a 1500 byte packet (1 MTU). So a 5 ms target should cause drops to occur. So increasing the target to 16ms with interval 111ms would be more appropriate for your upload speed. On the other hand you're working just fine with Traditional QoS with 5ms/100ms.

I will take a deeper look this weekend.

 

[BoostOver]

Please post or PM me the contents of these commands when running Traditional QoS. You might need to upload it to a site like pastebin.com if it's too much data to post.

Report
I would also like to remind you that bufferbloat (download) is better in FlexQoS.
Thanks again

 

[dave14305]

Report
I would also like to remind you that bufferbloat (download) is better in FlexQoS.
Thanks again

I’m still very puzzled by the output. It seems like you’re missing several upload classes, such as the default 1:40 class. Maybe Traditional QoS isn’t supposed to be setup on ppp0, but on eth0. Looks broken.

So I will take the results with a grain of salt.

Maybe run this to see if errors are produced:

/tmp/qos start

 

[BoostOver]

Sono ancora molto perplesso dall'output. Sembra che manchino diverse classi di caricamento, come la classe 1:40 predefinita. Forse la QoS tradizionale non dovrebbe essere impostata su ppp0, ma su eth0. Sembra rotto.

Quindi prenderò i risultati con le pinze.

Forse eseguilo per vedere se vengono prodotti errori:
[CODE = bash] / tmp / qos start [/ CODE]

ASUSWRT-Merlin DSL-AC68U 386.01_2-gnuton0_beta2 ven 19 feb 22:33:43 UTC 2021
Nutella05 @ DSL-AC68U-AF00: / tmp / home / root # tmp / qos start
-sh: tmp / qos: non trovato
Nutella05 @ DSL-AC68U-AF00: / tmp / home / root #

 

[sentinelvdx]

@dave14305 does Flex works with dual wan?
I'm looking to enable dual wan with a USB 4G Dongle for when my main ISP goes down.

Thanks

 

[dave14305]

@dave14305 does Flex works with dual wan?
I'm looking to enable dual wan with a USB 4G Dongle for when my main ISP goes down.

Thanks

In theory it does. @chris.at has a dual-WAN setup I believe.

 

[dave14305]

ASUSWRT-Merlin DSL-AC68U 386.01_2-gnuton0_beta2 ven 19 feb 22:33:43 UTC 2021
Nutella05 @ DSL-AC68U-AF00: / tmp / home / root # tmp / qos start
-sh: tmp / qos: non trovato
Nutella05 @ DSL-AC68U-AF00: / tmp / home / root #

Translator added unnecessary spaces. See original post again.

EDIT: Not necessary. Traditional QoS is certainly broke in 386.1 based on the output you’ve shown. There is no upload control being applied anymore (no classes 1:10 - 1:40 on the WAN interface). It looks like a code merge issue with ASUS GPL 41700, but it doesn’t deter my plans. But remember that Traditional QoS is not properly limiting upload anymore when comparing results with FlexQoS.