FlexQoS FlexQoS 1.0 - Flexible QoS Enhancement Script for Adaptive QoS

[dave14305]

Nothing wrong @dave14305 but wanted to bring this to your attention. Running flexqos -update when already on latest version shows this. No biggy in the grand scheme of things.
View attachment 24298

I'm still trying to come up with a good versioning plan. I only plan to increment the version number at particular milestones, but still commit various "hotfixes" or incremental updates in between, which the md5 comparison is detecting. So I need some better language in the prompt. The responsible thing to do would be to limit my ad-hoc commits to a separate develop branch. Writing code was so much simpler before there were people using it.

 

[Adamm]

I'm still trying to come up with a good versioning plan. I only plan to increment the version number at particular milestones, but still commit various "hotfixes" or incremental updates in between, which the md5 comparison is detecting. So I need some better language in the prompt. The responsible thing to do would be to limit my ad-hoc commits to a separate develop branch. Writing code was so much simpler before there were people using it.

Here's some wisdom, versioning systems on scripts like these never play out how you expect them too I also suggest a 3 decimal versioning system as you will run out of numbers much quicker then anticipated

 

[QuikSilver]

Here's some wisdom, versioning systems on scripts like these never play out how you expect them too I also suggest a 3 decimal versioning system as you will run out of numbers much quicker then anticipated

Wise words from @Adamm indeed. With the old method we will be on v1.0 in no time.

 

[dave14305]

Version 0.7.0 beta is posted.

Changes:

• Fix burst and cburst detection
• Fix word-wrap with long device names
• Move tc rule counting outside of iptables if structure to fix -check behavior
• Reduce expected tc filter rule count to aid AX users or 384.17 users
• Prevent invalid AppDB rules from being rendered in table
• Trigger -check in the background after a wrs restart (sig update)
• Cleanup some of the logging verbosity
• Convert the Customize toggle text to a button
• Adjust layout of AppDB table to avoid wrapping the buttons on iOS Safari
• Fix the http URL shown after install/update/help (@QuikSilver)
  
• Make all yes/no prompts have consistent numeric responses (@maghuro)

flexqos -update

 

[Makaveli]

Why kind of improvements are you expecting? I wouldn't think you would with those kinds of speeds. Most have lower speeds and want items like VOIP and video to be a higher priority than a file download or gaming. With your speeds I wouldn't think those items would get maxed out or bottlenecked.

The only place I could see this being helpful is with downloading. My connection is similar in speeds and if you have 1 machine that can pull down the full bandwidth of the connection you could potential see some issues with others streams on the network. Key word is "may" and it would have to be tested. For the most people anyone on 1Gbps+ fiber connection shouldn't need QOS on.

 

[lilstone87]

Should I be seeing the two new categories on my AX88U on the QoS page? I am running Merlin's 384.18_beta1. I currently don't see Work-from-home or Learn-from-home.

EDIT: From Merlin's change log, he used GPL 384_8563 for the AX models. Which supposedly Asus included these two new categories in their "8018" build. So I find it odd, why I am not seeing these two categories on the QoS page.

 

[L&LD]

No.

 

[lilstone87]

No.

I am curious why not? Isn't Merlin's code for this beta newer than what was included with the "8018" build for the AX88U, that Asus released that was said to include these two new categories?

EDIT: I see at least on Asus site, they never released the source code for the 8018 build for the AX88U

 

[dave14305]

I am curious why not? Isn't Merlin's code for this beta newer than what was included with the "8018" build for the AX88U, that Asus released that was said to include these two new categories?

EDIT: I see at least on Asus site, they never released the source code for the 8018 build for the AX88U

I'm out of my league in trying to unravel the AX issues, but I think the AX88U in 384.18b1 still has binary blobs from 384_7977.

 

[RMerlin]

I am curious why not? Isn't Merlin's code for this beta newer than what was included with the "8018" build for the AX88U, that Asus released that was said to include these two new categories?

EDIT: I see at least on Asus site, they never released the source code for the 8018 build for the AX88U

Right, and the newer GPL I have for the RT-AX88U is not compatible with what's currently used for both the AX56 and AX58, so it won't be merged until 384.19.

 

[lilstone87]

Right, and the newer GPL I have for the RT-AX88U is not compatible with what's currently used for both the AX56 and AX58, so it won't be merged until 384.19.

@RMerlin I appreciate you taking the time to clear this up, and responding here. Until this change happens, things are a bit messy on the AX88U at least on the FlexQoS script.

 

[RMerlin]

@RMerlin I appreciate you taking the time to clear this up, and responding here. Until this change happens, things are a bit messy on the AX88U at least on the FlexQoS script.

And the sad thing is, these two new categories are completely useless. One is just Messenging that was renamed, and the other one reuses two traffic classes already in other categories,, causing problems like some web traffic being incorrectly classified as people reported.

More problematic for this script is the fact that these changes were also tied to other changes in the available prios. The insertion of Learn-From-Home shifted prios down by one, and forced the removal of the old Default category.

Not sure what Trend Micro had in mind when they changed this...

 

[lilstone87]

And the sad thing is, these two new categories are completely useless. One is just Messenging that was renamed, and the other one reuses two traffic classes already in other categories,, causing problems like some web traffic being incorrectly classified as people reported.

More problematic for this script is the fact that these changes were also tied to other changes in the available prios. The insertion of Learn-From-Home shifted prios down by one, and forced the removal of the old Default category.

Not sure what Trend Micro had in mind when they changed this...

Yeah overall silly on their part. I noticed what you're saying, as the Classification, and FlexQoS page, has traffic all over the place now. That's why I said it's messy. I honestly just wish they would of put a little more focus on underlining type issues with their QoS. The reason you had a person like FreshJR making a script, to make QoS run better.

Lastly.. a bit off topic, but I am curious for the routers that can run it, will it be possible to use something like "Cake" with Adaptive QoS in the future? I know Cake is very early right now, and very limited on router support. Still it's something cool to think about the possibility of it.

 

[lilstone87]

Hey @dave14305

I just wanted you to know, I appreciate you taking on this project. I previously used FreshJR's script for a while in the past. Which had a couple bugs, and issues. Which we thought would get fixed at some point. But as you know, he disappeared. I hope for his sake, it was his decision, and not something bad. As it was a bit odd how he disappeared, and been gone for a while now.

 

[RMerlin]

Lastly.. a bit off topic, but I am curious for the routers that can run it, will it be possible to use something like "Cake" with Adaptive QoS in the future?

No. Unlike fq_codel, cake isn't just a drop-in replacement for sfq, so to implement it would require Trend Micro to do it within their proprietary engine.

 

[dave14305]

Hey @dave14305

I just wanted you to know, I appreciate you taking on this project. I previously used FreshJR's script for a while in the past. Which had a couple bugs, and issues. Which we thought would get fixed at some point. But as you know, he disappeared. I hope for his sake, it was his decision, and not something bad. As it was a bit odd how he disappeared, and been gone for a while now.

Thanks. I saw this as a way to give back to the community that I have learned so much from. Heck, I wasn’t even using QoS or Merlin firmware when I saw several posts in the FreshJR thread wondering about its future. I was very content running John’s fork wide-open (sans QoS). But sometimes you have to go where the action is and break a few eggs, or any other suitable metaphor.

The biggest kick for me is figuring out how to think like a developer again. I learned COBOL, Pascal and HyperTalk in college, did SAP ABAP/4 in the late 90s, and wrote my first HTML page in February 2020. Middle-age dog, new tricks!

 

[dave14305]

More problematic for this script is the fact that these changes were also tied to other changes in the available prios. The insertion of Learn-From-Home shifted prios down by one, and forced the removal of the old Default category.

I’ve reinstated a default rule for unmarked traffic which doesn’t exist in stock at all with the new categories. I keep meaning to PM Vanic to see if I’m misreading things.

But now that Learn From Home is adjustable in the priority list, I was able to undo some of FreshJR’s forced manipulation of the old Defaults class that used to be stuck in last priority. As long as Learn-From-Home stays at the bottom, interference should be at a minimum.

 

[lilstone87]

No. Unlike fq_codel, cake isn't just a drop-in replacement for sfq, so to implement it would require Trend Micro to do it within their proprietary engine.

@RMerlin I just posted in the other thread about this, before seeing this post of yours. I appreciate this info, I trust you, and your work, over the years. Overall that sucks to hear. As having Cake as an option would be great, and helpful overall, to many people.

I tested both CakeQoS, and FlexQoS, for a couple hours today. I ran several fast.com tests. With CakeQoS, it kept loaded upstream latency between 16-22ms. With FlexQoS using FQ_Codel, running the same test several times. upstream loaded latency averaged between 56-62ms. Which is a decent sized jump, between the two. For me, I love having more options, with details, and info available. Which FlexQoS offers with Adaptive QoS.

 

[lilstone87]

This isn't directed at FlexQoS. But I have noticed in the last hour, or so, my phone which is connected to the routers wifi. When running a speedtest, it doesn't seem the upstream is being limited. Adaptive QoS set at 30, speedtest hitting 35, which is around my max from my isp.

I remember this being an issue a little while back.. did Asus never fix this? If so.. what the hell. QoS is a bit pointless if wifi traffic can just bypass upstream limits. @RMerlin wanted to see if you have any info on this.

 

[L&LD]

@lilstone87, you need to set it lower.