Is QoS broken?

[Simon W]

Is QoS broken in the latest firmware, at least for the RT-AX88U in my case? (latest firmware = @RMerlin 386.16_B1)

I've tested three use-cases, each using speedtest.net:

1. Upload/download from a wired client
2. Upload/download from a Wi-Fi connected client
3. Upload/download from the router itself (eg. spdMerlin)

TL;DR - in terms of QoS doing its thing to restrain traffic:

1. Wired: Upload = works, download = works
2. Wi-Fi: Upload = doesn't work, download = probably works
3. Router: Upload = works, download = doesn't work

Also, I note when enabling/disabling QoS via the web GUI the change applies to my testing straight away, whereas changes to the manual bandwidth limits, either on their own or in combination with an enable/disable, require a reboot to take effect. Is this a known thing? If so, is there a way to shortcut via a service restart or similar?

Detail

My shiny new connection, unrestrained (ie. QoS disabled), tests at: down=306Mbs, up=41Mbs. I've therefore used the 95% logic and set manual bandwidth to down=290.7 and up=38.95 (unnecessarily precise I know!) using Adaptive and fq_codel. I've left the WAN packet overhead as "0" because I don't know exactly what setting I should be using (Asus router double-NATing to an ISP router, in turn with a PPPoE over G.Fast WAN?) but also didn't want to confuse the results further!

My speedtest.net rounded results, which are very consistent (bar Wi-Fi), as follows:

1. Wired: down=282Mbs, up=38Mbs <-- seems to work
2. Wi-Fi: down= around 260Mbs, up=41Mbs <-- uploads definitely unrestrained, downloads probably a combination of QoS working and Wi-Fi congestion
3. Router (spdMerlin): down=306Mbs, up=38Mbs <-- downloads definitely unrestrained

Does anyone have QoS currently known to be working (across all three use-cases above)? Either on the AX88U, another router and/or another firmware version? I'm keen to know if this is a problem with my configuration, Merlin firmware, specific router models, or across the board..?

 

[Adamm]

TL;DR:

1. Wired: Upload = works, download = works
2. Wi-Fi: Upload = doesn't work, download = probably works
3. Router: Upload = works, download = doesn't work

This occurred during the WPA3 GPL merge, I guess it unfortunately still hasn't been fixed upstream.

 

[Simon W]

Thanks for the confirmation, at least I know now. In that case I'm now eagerly awaiting the next Asus firmware release, to then in turn be incorporated into a Merlin release!

 

[S.claus]

@axaro1 brought up the same issue yesterday but at this point, I wish they would give merlin full access so he can fix bugs like this inside the closed source code

 

[axaro1]

Adaptive QoS is definitely broken on my RT-AX56U with 386.16_B.
QoS in upload doesn't work whether I'm on wifi/cable, download seems to work properly.
I did a 'dirty' update when flashing to the 386.16 from 3.0.0.4.384.8253, what firmware update method did you use @Simon W ?Did you fully reset your router?

 

[Vexira]

Is QoS broken in the latest firmware, at least for the RT-AX88U in my case? (latest firmware = @RMerlin 386.16_B1)

I've tested three use-cases, each using speedtest.net:

1. Upload/download from a wired client
2. Upload/download from a Wi-Fi connected client
3. Upload/download from the router itself (eg. spdMerlin)

TL;DR - in terms of QoS doing its thing to restrain traffic:

1. Wired: Upload = works, download = works
2. Wi-Fi: Upload = doesn't work, download = probably works
3. Router: Upload = works, download = doesn't work

Also, I note when enabling/disabling QoS via the web GUI the change applies to my testing straight away, whereas changes to the manual bandwidth limits, either on their own or in combination with an enable/disable, require a reboot to take effect. Is this a known thing? If so, is there a way to shortcut via a service restart or similar?

Detail

My shiny new connection, unrestrained (ie. QoS disabled), tests at: down=306Mbs, up=41Mbs. I've therefore used the 95% logic and set manual bandwidth to down=290.7 and up=38.95 (unnecessarily precise I know!) using Adaptive and fq_codel. I've left the WAN packet overhead as "0" because I don't know exactly what setting I should be using (Asus router double-NATing to an ISP router, in turn with a PPPoE over G.Fast WAN?) but also didn't want to confuse the results further!

My speedtest.net rounded results, which are very consistent (bar Wi-Fi), as follows:

1. Wired: down=282Mbs, up=38Mbs <-- seems to work
2. Wi-Fi: down= around 260Mbs, up=41Mbs <-- uploads definitely unrestrained, downloads probably a combination of QoS working and Wi-Fi congestion
3. Router (spdMerlin): down=306Mbs, up=38Mbs <-- downloads definitely unrestrained

Does anyone have QoS currently known to be working (across all three use-cases above)? Either on the AX88U, another router and/or another firmware version? I'm keen to know if this is a problem with my configuration, Merlin firmware, specific router models, or across the board..?

Gfast is VDSL2 just use the overhead for pppoe VDSL2 is close enough.

 

[S.claus]

Now it begs the question of does Adaptive QOS actually prioritize the right packets (like let’s say game packets ) ?

 

[Vexira]

Now it begs the question of does Adaptive QOS actually prioritize the right packets (like let’s say game packets ) ?

It does if the traffic has been added to the database in a signature update.

If it hasn't then no.

 

[Simon W]

Adaptive QoS is definitely broken on my RT-AX56U with 386.16_B.
QoS in upload doesn't work whether I'm on wifi/cable, download seems to work properly.
I did a 'dirty' update when flashing to the 386.16 from 3.0.0.4.384.8253, what firmware update method did you use @Simon W ?Did you fully reset your router?

Definitely did a full reset and manual configuration. For one reason or another, not related to QoS I may add, I've done this around 8 times over the past week, have got it down to a fine art!

 

[axaro1]

Now it begs the question of does Adaptive QOS actually prioritize the right packets (like let’s say game packets ) ?

That's a good question, it is definitely detecting games like Overwatch and R6 as Gaming traffic but I don't know if it's effectively prioritizing these packets.
I can try to look at the network graph in Overwatch to confirm if the QoS is working.
*while pinging the server

 

[Simon W]

So while I have the attention of some QoS-knowledgable minds and as we wait patiently for a fix from Asus, a quick to aid my understanding of the implementation please:

When Adaptive QoS is fully working.. in a test environment, with only a single client connect via wired ethernet (ie. removing as many variables as possible), if I set my bandwidth limits to down=290.7 and up=38.95, along with getting my WAN packet overhead correct (thanks @Vexira!) should I expect to see matching results from speedtest.net? Or, as per my testing above in the cases where it appears to be working, should I see a single client reaching "near to" these figures (my results seem to be around 97%)?

FYI. having moved from a real-world 60/17 connection to now over 300/40.. I'm very content with sacrificing some Mbs in exchange for a reliable bandwidth sharable across the network.. I just wanting to satisfy my curiosity of how these settings work thanks.

 

[Kingp1n]

So while I have the attention of some QoS-knowledgable minds and as we wait patiently for a fix from Asus, a quick to aid my understanding of the implementation please:

When Adaptive QoS is fully working.. in a test environment, with only a single client connect via wired ethernet (ie. removing as many variables as possible), if I set my bandwidth limits to down=290.7 and up=38.95, along with getting my WAN packet overhead correct (thanks @Vexira!) should I expect to see matching results from speedtest.net? Or, as per my testing above in the cases where it appears to be working, should I see a single client reaching "near to" these figures (my results seem to be around 97%)?

FYI. having moved from a real-world 60/17 connection to now over 300/40.. I'm very content with sacrificing some Mbs in exchange for a reliable bandwidth sharable across the network.. I just wanting to satisfy my curiosity of how these settings work thanks.

Have you tried testing with FreshJr script? I wonder if this would fix the A.QoS issue?

 

[S.claus]

Have you tried testing with FreshJr script? I wonder if this would fix the A.QoS issue?

Tried it a few weeks ago on my ax88u (200/100) and it didn’t fix any problems .

 

[Frankflash]

Qos needs to be upgraded to AQM with driver support

 

[Vexira]

So while I have the attention of some QoS-knowledgable minds and as we wait patiently for a fix from Asus, a quick to aid my understanding of the implementation please:

When Adaptive QoS is fully working.. in a test environment, with only a single client connect via wired ethernet (ie. removing as many variables as possible), if I set my bandwidth limits to down=290.7 and up=38.95, along with getting my WAN packet overhead correct (thanks @Vexira!) should I expect to see matching results from speedtest.net? Or, as per my testing above in the cases where it appears to be working, should I see a single client reaching "near to" these figures (my results seem to be around 97%)?

FYI. having moved from a real-world 60/17 connection to now over 300/40.. I'm very content with sacrificing some Mbs in exchange for a reliable bandwidth sharable across the network.. I just wanting to satisfy my curiosity of how these settings work thanks.

Speed test adjusts to the new settings, especially when you add the overhead, I can confirm form the last speed tests I did.

 

[Vexira]

Qos needs to be upgraded to AQM with driver support

I agree but we would have to ask Asus for that, probably have to create a thread in the Asus wrt section of the forum and have as many users as possible voice support.

 

[Frankflash]

i have sent asus feed back asking for it

 

[Vexira]

i have sent asus feed back asking for it

Awesome I'll look into doing the same.

 

[eclp]

Perhaps a very good way to get on with this problem would be Jack Cheng from ASUS Australia, who is omnipresent in the Australian Wirlpool forum.

For example, in the AX88U thread: https://forums.whirlpool.net.au/thread/3272x4m9?p=43

 

[Frankflash]

Perhaps a very good way to get on with this problem would be Jack Cheng from ASUS Australia, who is omnipresent in the Australian Wirlpool forum.

For example, in the AX88U thread: https://forums.whirlpool.net.au/thread/3272x4m9?p=43

all ready done that