FlexQoS FlexQoS 1.2.4 - Flexible QoS Enhancement Script for Adaptive QoS

[Skeptical.me]

I'm doing the same thing with Express VPN, sure it works, but you won't be able to optimize uploads. Go to the router's open vpn configuration page and in time put Accept DNS Configuration to Exclusive. Then go down to the advanced settings and force all the traffic through the tunnel with 'policy rules (strict)', further down select all your devices one by one and decide which one must pass in the VPN and which one not ... .... I have them all in VPN, then save everything. React the adaptive qos of the router and set custom priorities, put streaming etc first and put learn to home last.

Go to the qos flez and under customize select Router / VPN Client Outbound Traffic Class on Streaming

Thanks for the reply.

I have always used "Accept DNS Configuration" to "Exclusive" and forced all traffic through the VPN ... but without using Policy Rules (Strict). The reason for this is I've found that Diversion won't work. I see adverts everywhere. So, I set "force all the traffic through the tunnel" to "Yes". Then Diversion works.

 

[MatteoPV]

Thanks for the reply.

I have always used "Accept DNS Configuration" to "Exclusive" and forced all traffic through the VPN ... but without using Policy Rules (Strict). The reason for this is I've found that Diversion won't work. I see adverts everywhere. So, I set "force all the traffic through the tunnel" to "Yes". Then Diversion works.

You're right, it works !!!!! Are you also using Unbound?

 

[Skeptical.me]

If you do this, all traffic goes through the tunnel and not the VPN, so everything except the VPN will work. With your setting all traffic goes through the tunnel and avoids the vpn

oh, okay.

I was with the understanding that "force all traffic through the tunnel: Yes" meant you force all traffic through the VPN tunnel.

I mean it must be working because I can stream HULU, Netflix US, Amazon Prime US, and all the cable TV Apps (I have access to a DirecTV account) from Australia. And when I stop the VPN I can't watch any of those streaming apps (but I can stream the Australian versions of Netflix and Amazon Prime).

Also, when I go to IPleak.net, as well as this site https://www.expressvpn.com/dns-leak-test the tests show I'm connected to ExpressVPN servers.

 

[MatteoPV]

oh, okay.

I was with the understanding that "force all traffic through the tunnel: Yes" meant you force all traffic through the VPN tunnel.

I mean it must be working because I can stream HULU, Netflix US, Amazon Prime US, and all the cable TV Apps (I have access to a DirecTV account) from Australia. And when I stop the VPN I can't watch any of those streaming apps (but I can stream the Australian versions of Netflix and Amazon Prime).

Also, when I go to IPleak.net, as well as this site https://www.expressvpn.com/dns-leak-test the tests show I'm connected to ExpressVPN servers.

I edited my answer from earlier, you're right it works !!!! I set it up like this now too

 

[Skeptical.me]

N

You're right, it works !!!!! Are you also using Unbound?

No, because I'm not a 100% sure how unbound works. I have used it before but I saw that the expressvpn DNS servers weren't being used, but rather my IP address was being used as DNS when preforming a DNS test. But we better keep on topic in this thread

 

[Skeptical.me]

Should I disable the QoS in the Ui and only use FlexQoS? Or should both be enabled?

 

[nikr]

I just installed it yesterday, and if I remember correctly it specifically asked me to enable QoS in UI before proceeding with install.

 

[immediatesalt]

Hello everyone!

Firstly, thanks Merlin for your awesome efforts on this firmware! I have used it on many various devices over the years, even though is my first post at SNB.

My question is regarding bufferbloat. I have been utilizing DSLReports' speedtest and it consistent tells me that my bufferbloat score is C, D, or F. I have tried quite a few options in the QOS and also installed FlexQOS. However, I can't seem to get the bufferbloat to improve, I've even cut my bandwidth settings down to 75% of inbound/outbound to 15/182.5 and there's no improvement.
With work-from-home and school-from-home streams running, we are finding that internet responsiveness is a bit of a struggle.

Is there anyone here that could share their experience on improving bufferbloat?

I have an ASUS AC3100 with two ASUS AC68Us in an AI mesh setup. All units have been upgraded to the latest 386.2_2 Asuswrt-Merlin firmware, and FlexQOS has been updated to 1.2.4.

Thanks!

 

[geobernd]

However, I can't seem to get the bufferbloat to improve, I've even cut my bandwidth settings down to 75% of inbound/outbound to 15/182.5 and there's no improvement.

Did you by any chance switch inbound and outbound (the way you posted it shows you have 15 mpbs inbound and 182.5 mbps outbound - which is probably the wrong way around)....?

 

[immediatesalt]

Did you by any chance switch inbound and outbound (the way you posted it shows you have 15 mpbs inbound and 182.5 mbps outbound - which is probably the wrong way around)....?

Oh yes, absolutely you are correct. I screwed up above - I actually meant 182.5 download / 15 upload.

 

[immediatesalt]

Also, on a tangential note, I have been adding some IPtables entries in FlexQOS. I was wondering if anyone knows of where to find a listing table of default 'marks' for various applications?

It seems like traffic is marked by TrendMicro as part of the AdaptiveQOS algorithm? However, I am unable to find any table of typical mark codes, so I have to manually wait or create specific traffic to figure out the mark codes in the FlexQOS screens.

Primarily, I'm trying to prioritize Webex, Teams, Google Meet, and Zoom into Work-from-Home correctly. It seems like a bunch of them are misclassified into STUN, when ends up in my Other category, and I'd like to fix that.

 

[dave14305]

I was wondering if anyone knows of where to find a listing table of default 'marks' for various applications?

You can search in the AppDB Application name field.

Primarily, I'm trying to prioritize Webex, Teams, Google Meet, and Zoom into Work-from-Home correctly. It seems like a bunch of them are misclassified into STUN, when ends up in my Other category, and I'd like to fix that.

If they end up as STUN, then create an AppDB rule to put STUN into Work-From-Home.

 

[immediatesalt]

Thank you! That was very helpful.
Out of curiosity - would you be able to explain the difference between the IPTable and AppDB rule listings?
- IPTable = classification by port/IP/mark
- AppDB = classification by mark
Is that basically it?

For what it's worth, even after adding the lines to the IPTable and AppDB, it does not seem to improve my DSLReports bufferbloat ratings - still C or worse.

 

[dave14305]

Out of curiosity - would you be able to explain the difference between the IPTable and AppDB rule listings?
- IPTable = classification by port/IP/mark
- AppDB = classification by mark
Is that basically it?

Use an iptables rule when you want to override what Application the traffic is classified as (or it’s Untracked). Use AppDB rules when you don’t like what class an application is assigned to (Work-From-Home, Gaming, Web Surfing).

For what it's worth, even after adding the lines to the IPTable and AppDB, it does not seem to improve my DSLReports bufferbloat ratings - still C or worse.

Rules won’t impact bufferbloat. Your configured speed and overhead values will most affect your bufferbloat. Keep reducing speeds until it starts to get better.

 

[immediatesalt]

So bufferbloat is purely a function of speed?

So I ran a battery of tests with both DSL Reports and Flent against netperf-west.bufferbloat.net (netperf-east was down) in a Ubuntu VM over a wifi connection, and it seems that I don't get good bufferbloat results until I scale things down considerably.

I have a 250/20 package. I don't get good bufferbloat scores until I scale my QOS bandwidth numbers to 100/20.

My internet line is cable with Start Communications in Toronto. Router is Asus AC3100 with latest Merlin and FlexQOS.
Is that expected with a setup like mine?

 

[dave14305]

So bufferbloat is purely a function of speed?

No, but your speeds must be set low enough to ensure that your router IS the bottleneck and not your upstream ISP. If your actual ISP line speed falls below your configured QoS speeds, you’re no longer in control.

So I ran a battery of tests with both DSL Reports and Flent against netperf-west.bufferbloat.net (netperf-east was down) in a Ubuntu VM

You’re pretty hardcore if you’re running Flent.

I have a 250/20 package. I don't get good bufferbloat scores until I scale my QOS bandwidth numbers to 100/20.

Why not lower the upload to 18? You’re really only fully in charge of data you send toward the Internet (uploads). Throttling downloads means the bits have already consumed your ISP link to get to your router to be throttled. Spilled milk.

 

[immediatesalt]

No, but your speeds must be set low enough to ensure that your router IS the bottleneck and not your upstream ISP. If your actual ISP line speed falls below your configured QoS speeds, you’re no longer in control.

Line speed doesn't seem to be an issue at the moment. All my raw speed tests at all times of day are showing 270/20 consistently.

You’re pretty hardcore if you’re running Flent.

I wouldn't call myself hardcore by any stretch of the imagination!

Why not lower the upload to 18? You’re really only fully in charge of data you send toward the Internet (uploads). Throttling downloads means the bits have already consumed your ISP link to get to your router to be throttled. Spilled milk.

Well, my testing yesterday afternoon (4pm) on a quiet network (I kicked the kids off for 45 mins) showed that upload speed from 10 up to 20 had no effect - only saw changes when I put the setting at 25. Download speed, however, did affect latency/bufferbloat from 250 all the way down until I came to 100.

That said, this morning (6am) testing at the same speeds showed worse results. The same 100/20 speeds that gave me an A on DSLReports yesterday are giving me C and D results and latencies are all over the place again. Aaaagh! What is going on? I would have expected the results to be more stable, seeing as the server I'm hitting netperf-west.bufferbloat.net should be quiet (3am).

I don't know what to make of it. I'm super frustrated. I'm questioning my tests. I'm questioning bufferbloat. I'm questioning how well ASUS QOS works. I'm questioning if FlexQOS does anything to help my overall internet experience. None of this data makes sense to me. I'm throwing my hands in the air and setting my speeds to 225/18 - which probably means I'll spend 24 hours searching the internet for more information and will revisit this all tomorrow. Sigh.

 

[dave14305]

I don't know what to make of it. I'm super frustrated. I'm questioning my tests. I'm questioning bufferbloat. I'm questioning how well ASUS QOS works. I'm questioning if FlexQOS does anything to help my overall internet experience. None of this data makes sense to me. I'm throwing my hands in the air and setting my speeds to 225/18 - which probably means I'll spend 24 hours searching the internet for more information and will revisit this all tomorrow. Sigh.

Sometimes when nothing seems to help, it can mean your problem is outside your network and therefore out of your control.

 

[immediatesalt]

Sometimes when nothing seems to help, it can mean your problem is outside your network and therefore out of your control.

Thanks for this observation. It’s helpful for my sanity.

 

[docious]

Hello everyone!

Firstly, thanks Merlin for your awesome efforts on this firmware! I have used it on many various devices over the years, even though is my first post at SNB.

My question is regarding bufferbloat. I have been utilizing DSLReports' speedtest and it consistent tells me that my bufferbloat score is C, D, or F. I have tried quite a few options in the QOS and also installed FlexQOS. However, I can't seem to get the bufferbloat to improve, I've even cut my bandwidth settings down to 75% of inbound/outbound to 15/182.5 and there's no improvement.
With work-from-home and school-from-home streams running, we are finding that internet responsiveness is a bit of a struggle.

Is there anyone here that could share their experience on improving bufferbloat?

I have an ASUS AC3100 with two ASUS AC68Us in an AI mesh setup. All units have been upgraded to the latest 386.2_2 Asuswrt-Merlin firmware, and FlexQOS has been updated to 1.2.4.

Thanks!

Forgive me for even suggesting it but are you testing for bufferbloat while connected to a VPN? I felt very stupid after realising I was doing that with ExpressVPN. My DSLreports bufferbloat rating was stuck on C no matter what I did with QoS. With the VPN disconnected and the rating became A. Not my best moment!