Adaptive QoS Broken

[Find the Door]

I believe there's something wrong with Adaptive QoS on 384.4 _2 (AC86U). With the same values I get the below results with consistency. I have a Netgear CM600 hooked up currently but get similar results with a CM1000 or SB8200 hooked up as well. This is using fq_codel with the same up and download values entered with 18 wan packet overhead (still happens regardless if set at 0).

Download will preform flawlessly - upload does the following however

Adaptive QoS

Traditional QoS

When I had my RT-AC5300 running Brainslayers DDWRT fq_codel didn't do this at all - it was essentially set it and forget it. Before anyone goes posting "try FreshJr's QOS it's gold" I'm not saying it isn't. This happens with and without his script running so clearly something else is at play here. I'd like to be able to use adaptive for obvious reasons.

 

[Sinner]

I believe there's something wrong with Adaptive QoS on 384.4 _2 (AC86U). With the same values I get the below results with consistency. I have a Netgear CM600 hooked up currently but get similar results with a CM1000 or SB8200 hooked up as well. This is using fq_codel with the same up and download values entered with 18 wan packet overhead (still happens regardless if set at 0).

Download will preform flawlessly - upload does the following however

Adaptive QoS

Traditional QoS

When I had my RT-AC5300 running Brainslayers DDWRT fq_codel didn't do this at all - it was essentially set it and forget it. Before anyone goes posting "try FreshJr's QOS it's gold" I'm not saying it isn't. This happens with and without his script running so clearly something else is at play here. I'd like to be able to use adaptive for obvious reasons.

I have a similar effect on my uploads regardless of what i set it at. No qos i can usually get over 4mbs with bloat. With qos regarless of what i set it at all the way down to 0.8 i get the same thing but... it appears to be working well and bloat was much improved. If i recall we have a limited backported version of fq_codel so it may be related to that.

 

[Find the Door]

I have a similar effect on my uploads regardless of what i set it at. No qos i can usually get over 4mbs with bloat. With qos regarless of what i set it at all the way down to 0.8 i get the same thing but... it appears to be working well and bloat was much improved. If i recall we have a limited backported version of fq_codel so it may be related to that.

Yeah, when I was on Brainslayers DDWRT build FQ_Codel was bulletproof and performed perfectly - I was floored! It's too bad my RT-AC5300 took a dump as now with the AC86U I don't have much choice but to use stock or Merlin.

 

[Find the Door]

Is there any chance this could be updated or something like cake could be added?


Because as of now it appears to only be functioning on the download side of things.

 

[NGRhodes]

Have you tried the SQF Queue Discipline instead of FQ_Codel with adaptive QoS on ?

 

[RMerlin]

r something like cake could be added?

No. Adaptive QoS is closed source, cake would require access to that code to completely redesign the rules it generates.

 

[Find the Door]

Have you tried the SQF Queue Discipline instead of FQ_Codel with adaptive QoS on ?

Yeah I've tried SFQ but it's nearly the same. A little better but not by much.

 

[Find the Door]

No. Adaptive QoS is closed source, cake would require access to that code to completely redesign the rules it generates.

Darn that's unfortunate - a more updated version of fq_codel perhaps?

 

[Sinner]

I did a bunch of reading on the differences between fq_codel etc. Unless your a gamer your probably further ahead with sfq on asus products. The only real advantage to fq_codel is how it sorta proritizes smaller packets 1st like those often used by online gaming because they are very small amounts of data but very timing aensitive.

 

[Find the Door]

I did a bunch of reading on the differences between fq_codel etc. Unless your a gamer your probably further ahead with sfq on asus products. The only real advantage to fq_codel is how it sorta proritizes smaller packets 1st like those often used by online gaming because they are very small amounts of data but very timing aensitive.

That I am and thus why I want to make use of it

 

[Sinner]

That I am and thus why I want to make use of it

Well it works decent imo for gaming with freshjr. Ive tried and cant gain more than 5-15 ping nomatter what i do. Im quite happy with that vs 3000+ that i could be getting doing the same things.

I have mobile cell internet. With qos/freshjr usually about 80 ping and no more than 91-97 under max load. Works for me

 

[Find the Door]

I just don't understand why it is that Asus' QoS implementation needs so much doctoring. You shouldn't have to install 3rd party firmware in the first place to get a semi-functional build.

I've since search dozens of threads to find similar results with fq_codel. Kind of regretting the 86U purchase at this point.

 

[RMerlin]

just don't understand why it is that Asus' QoS implementation needs so much doctoring. You shouldn't have to install 3rd party firmware in the first place to get a semi-functional build.

Part of the problem is there is no real one-size-fit-all solution when it comes to QoS. Asus' implementation was designed to be as easy as possible for novice users to use, which is why it has limited configuration settings, and for many users it works properly. I know in my own case it gets the job done. Some users might have very specific requirements however, in which case you need something more advanced - that's beyond the scope of what Asuwrt is intended to provide.

 

[mromero]

[QUOTE="RMerlin, post: 395019, member: 10954" I know in my own case it gets the job done. [/QUOTE]

What in your case are you using? Traditional QOS?

Asus Adaptive QOS is busted and shame on a big company pushing out broken firmware on the AC 86U is my judgement.

I will not be surprised to find out Trend Micro, a Taiwanese company, is also compromised.

Reason we still run a commercial grade hardware firewall with traffic shaper in front just in case....

 

[Find the Door]

[QUOTE="RMerlin, post: 395019, member: 10954" I know in my own case it gets the job done.

What in your case are you using? Traditional QOS?

Asus Adaptive QOS is busted and shame on a big company pushing out broken firmware on the AC 86U is my judgement.

I will not be surprised to find out Trend Micro, a Taiwanese company, is also compromised.

Reason we still run a commercial grade hardware firewall with traffic shaper in front just in case....[/QUOTE]I was trying to run adaptive - but unfortunately it does seem to be busted. I don't know how it could be functioning properly for other users when it's giving me the off the wall results it is. I guess they're just not as keen into optimizing their connections?

 

[RMerlin]

What in your case are you using? Traditional QOS?

I'm using what this thread is discussing - Adaptive QoS.

 

[skeal]

I use FreshJR_QOS and it works awesome as a Adaptive QOS in concert with Asus's QOS (fq_codel of course). This works real well and classifies traffic well when all setup properly. If you need more granular control you can get devices made for that purpose. This is a well supported home and small office grade router nothing more nothing less.

 

[Find the Door]

I use FreshJR_QOS and it works awesome as a Adaptive QOS in concert with Asus's QOS (fq_codel of course). This works real well and classifies traffic well when all setup properly. If you need more granular control you can get devices made for that purpose. This is a well supported home and small office grade router nothing more nothing less.

I understand, but something like that XR500 that features Duma OS completely trump's our current implementation. I don't think that should be the case especially after what I briefly experienced with Brainslayers DDWRT build. Fq_codel completely remedied any and all lag spikes I was experiencing on his firmware. No matter if I had started a 4k HDR stream or not - it was marvelous.


I don't think it's unreasonable to expect a working, functional but above all else effective implementation of QoS in 2018.