FlexQoS FlexQoS 1.0 - Flexible QoS Enhancement Script for Adaptive QoS

[Jack Yaz]

It’s been a fun time for me since the Addon API was released by RMerlin in February. By February 13 I had created my first UI for Unbound, and once the pandemic shut things down, there was plenty of time to dig deeper. In early June I started playing with FreshJR_QOS. And here I am today.

I am truly impressed by how extensible the Merlin firmware is. It’s quite ingenious how all the “hooks” are designed and how well they work for most use cases.

I don't miss scouring for an asp page I can override with one of my own pages, that's for certain!

 

[marko]

Hi, can anyone provide me the solution for this request.

I want to put ports 3478,3479, 3480, 5735, and many more to the gaming rule. And all except those to sort them to "others". Or to rephrase it... !3478, 3479, 3480... = "Others" and the rest in the gaming, for the specific IP address.

Solution? Thanks

 

[dave14305]

Hi, can anyone provide me the solution for this request.

I want to put ports 3478,3479, 3480, 5735, and many more to the gaming rule. And all except those to sort them to "others". Or to rephrase it... !3478, 3479, 3480... = "Others" and the rest in the gaming, for the specific IP address.

Solution? Thanks

How many ports, exactly? There are limits within both FlexQoS and iptables on the number of individual ports. iptables has a limit of 15 ports. But FlexQoS has a limit of 36 characters for remote or local ports in the webui due to space constraints. There may be other ways to do what you want besides long port lists if you explain in more detail how it looks today and what you want it to look like in the future.

 

[marko]

How many ports, exactly? There are limits within both FlexQoS and iptables on the number of individual ports. iptables has a limit of 15 ports. But FlexQoS has a limit of 36 characters for remote or local ports in the webui due to space constraints. There may be other ways to do what you want besides long port lists if you explain in more detail how it looks today and what you want it to look like in the future.

I figured out that when playing P2P games there are a lot of IP connections and a lot of ports used... Generally i would then include only ports (that i would normally forward) as a gaming rules, while all other ports that are not those ones i would add them to the section others or some other section(game downloads for example).

Ports marked red are always changing (With starting of 1 game they are like this, they will change on the start of another game etc.) and i dont know what they are used for... port marked red is always there, and when searching for "port forward" for pes it is stated that this port should be forwarded, so i have set it for the gaming rule, along with some others (i am still testing them).

All other ports when included in the gaming rule slows down my game to really "impossible to play" speeds.

Edit: 9-10 ports lets say i would need for the gaming rule... There are 3 ports that are in range, for example 3478-3480

 

[dave14305]

I figured out that when playing P2P games there are a lot of IP connections and a lot of ports used... Generally i would then include only ports (that i would normally forward) as a gaming rules, while all other ports that are not those ones i would add them to the section others or some other section(game downloads for example).
View attachment 28100

Ports marked red are always changing (With starting of 1 game they are like this, they will change on the start of another game etc.) and i dont know what they are used for... port marked red is always there, and when searching for "port forward" for pes it is stated that this port should be forwarded, so i have set it for the gaming rule, along with some others (i am still testing them).

All other ports when included in the gaming rule slows down my game to really "impossible to play" speeds.

Edit: 9-10 ports lets say i would need for the gaming rule... There are 3 ports that are in range, for example 3478-3480

If I saw multiple udp connections while doing something that is latency sensitive, I would assume they are very important for the performance of what I'm doing (e.g. such as online meetings, VoIP, gaming). Perhaps another gamer can come up with a reasonable rule.

 

[marko]

Like i have stated, the most important port is 3478, unlike others. Other that might be important is STUN, which has its rule, but i have yet to test that.

So, after a short explanation, can you now provide me an answer for my question so i could test further? Thanks

 

[NGI]

Fresh meat on the develop branch.
...

Feedback appreciated!

View attachment 28097

Many thanks @dave14305 for another great addition, I love it. Perhaps another thing that would be very useful to me is to maybe have an indicator e.g. a horizontal bar showing the % of the total bandwidth utilization or the actual total bandwidth in use. Is that doable?

 

[dave14305]

Many thanks @dave14305 for another great addition, I love it. Perhaps another thing that would be very useful to me is to maybe have an indicator e.g. a horizontal bar showing the % of the total bandwidth utilization or the actual total bandwidth in use. Is that doable?

Don't mess up my momentum. I'm still working on optimizing the performance of the graph drawing since I reused Merlin's plumbing from the pie charts. I no longer .destroy() the graphs and redraw them, but now I just update them in-place.

Plus I'll have another side project later this afternoon and must give it the proper attention.

Spoiler

Should I mention the lackluster Black Friday/Cyber Monday sales on ASUS devices except the AX58U? I also have a $10 MIR form.

I don't fully understand your idea. Is it like the old Bandwidth Util bars I just removed, or something you want to see in the existing charts that would probably make them even more unreadable?

 

[NGI]

Is it like the old Bandwidth Util bars I just removed?

Yes but only one for the total bandwidth utilization as oppose to one for each Class and also outside of the charts. I tried to mock something below:

Ideally the speed not in MB/s but either Mbps or Kbps

 

[dave14305]

Yes but only one for the total bandwidth utilization as oppose to one for each Class and also outside of the charts. I tried to mock something below:

View attachment 28112


Ideally the speed not in MB/s but either Mbps or Kbps

I'm going to put that on the back burner since there's already a lot of drawing and redrawing going on in this page. While testing, I actually filter the tracked connection table with Proto=udp and Local Port=z to ensure it doesn't have to bother evaluating all the custom rules and connections to improve the responsiveness of the page when I'm ignoring that part of the page.

It's also a slipperly slope because as I mentioned earlier, the rate displayed on my page is a 10-second average that tc reports. The example graph you superimposed from the main page is a realtime data rate that Merlin calculates, which won't usually match the sum of the rates I'm displaying. But even if I use the sum of the class rates, there's still a possibility of confusing people less intelligent than yourself about what they're seeing.

Another reason to sleep on it is that I'm not very good about squeezing new html elements in tight spaces.

I'll add it to the Issues list on Github so I don't forget about it.

 

[NGI]

Thanks Dave, that makes sense.

 

[dave14305]

Update pushed to develop that hopefully improves performance a little bit and smooths out some rough edges from the initial attempt.

 

[JGrana]

Update pushed to develop that hopefully improves performance a little bit and smooths out some rough edges from the initial attempt.

Thanks for the update.

Now, go setup and enjoy your shiny new “toy” ( and I mean toy in a complimentary way )

Folks, look on the bright side - think of how much faster @dave14305 can debug. ;-)

 

[Kingp1n]

Don't mess up my momentum. I'm still working on optimizing the performance of the graph drawing since I reused Merlin's plumbing from the pie charts. I no longer .destroy() the graphs and redraw them, but now I just update them in-place.

Plus I'll have another side project later this afternoon and must give it the proper attention.

Spoiler

View attachment 28111
Should I mention the lackluster Black Friday/Cyber Monday sales on ASUS devices except the AX58U? I also have a $10 MIR form.

I don't fully understand your idea. Is it like the old Bandwidth Util bars I just removed, or something you want to see in the existing charts that would probably make them even more unreadable?

Maybe if time allows in the near future, would you maybe look into bringing cake to the GUI or possibly bring those nifty graphs along as well?
That would be an awesome accomplishment!!! Just me doing my early wish list before Xmas haha. Thanks @dave14305 for your handwork!

 

[Jack Yaz]

Don't mess up my momentum. I'm still working on optimizing the performance of the graph drawing since I reused Merlin's plumbing from the pie charts. I no longer .destroy() the graphs and redraw them, but now I just update them in-place.
?

i didn't see much of a performance hit with destroy and redraw vs. updating, but my charts are more of a draw once on page load kind of deal! some things I update or manually manipulate the options such as for my line annotations, chart fill etc. though where I did see a jump.

I'm waiting for chartjs v3 to go stable which promises massive performance gains, but it won't be a quick and dirty upgrade - a lot of migration needed due to breaking changes

 

[dave14305]

Maybe if time allows in the near future, would you maybe look into bringing cake to the GUI or possibly bring those nifty graphs along as well?
That would be an awesome accomplishment!!! Just me doing my early wish list before Xmas haha. Thanks @dave14305 for your handwork!

I will certainly be experimenting with CAKE once I setup the new router. It just arrived and it's still cold from the trip. I'm not sure how much there is to visualize with CAKE but time will tell.

 

[dave14305]

i didn't see much of a performance hit with destroy and redraw vs. updating, but my charts are more of a draw once on page load kind of deal! some things I update or manually manipulate the options such as for my line annotations, chart fill etc. though where I did see a jump.

I'm waiting for chartjs v3 to go stable which promises massive performance gains, but it won't be a quick and dirty upgrade - a lot of migration needed due to breaking changes

I'm more of an N-1 kind of guy. I only want to use what's baked into the firmware, if possible. You frontiersmen can blaze the trail for me.

 

[jata]

@dave14305 - big thanks for your help with flex and cake QoS queries. On the cake thread I have made some progress with getting it working on my AX58U / AX3000 but still not happy.

I have noticed that flexQoS does not limit my upload limit set in the QOS when using flexQoS - is this a known issue?

 

[dave14305]

@dave14305 - big thanks for your help with flex and cake QoS queries. On the cake thread I have made some progress with getting it working on my AX58U / AX3000 but still not happy.

I have noticed that flexQoS does not limit my upload limit set in the QOS when using flexQoS - is this a known issue?

There have been issues reported before with the AX58U and broken upload with stock Adaptive QoS.

 

[Milan]

Is there chance that the graphs can be in log scale ? And some check box on GUI to switch between normal and log scale ...