[Release] FreshJR Adaptive QOS (Improvements / Custom Rules / and Inner workings)

[maghuro]

Thanks. Is this with FreshJR disabled? I don’t see any filter rule for what used to be the undefined priority like below:

filter parent 1: protocol all pref 2 u32 fh 827::800 order 2048 key ht 827 bkt 0 flowid 1:17
  mark 0x80000000 0xc000ffff (success 8143)

Yes Dave, freshJR uninstalled as it would screw everything with the new categories.

As Merlin told on another thread, now the undefined category goes to the 4 one (the one you defined in the middle). Instead of going to the bottom

 

[Snipzy]

Yes, it still works and fq_codel is still the recommended setting. Remember to set your bandwidth speed manually and at 85-90% of your maximum measured speed to give QoS room to work. Measure your speed with DSLreports speed test or another tool that measures buffer bloat. IIRC the script takes the entered value and converts to kbps so you can use decimals, and in your case with only 0.5Mbps up, you'll want to go to hundredths, e.g 0.45 not just 0.4.

Test the settings via a speed test with buffer bloat and you'll see that QoS is working.

it's me again well it worked but i have a lot of bufferbloat around 2000ms stable i tried reducing the download and upload speed but it stay the same what can i do?

 

[nzwayne]

I think the first bullet is straightforward enough.

Having 4 and 13 present in multiple classes means having to rework the flowid determination since it uses 4 to determine the assigned priority for Streaming, and 13 to determine the priority for Web traffic. We can use category 12 for Web, but I don’t know yet what else would uniquely identify Streaming.

The elimination of the Default priority might mean the end of the Game Transferring custom class. Still wrapping my head around the implications.

But I think the default of untracked traffic going to priority 4 (1:14) is a belated win for FreshJR, but still thinking this through as well.

If anyone on 384.18 alpha wants to post/PM the output of these commands while FreshJR is disabled, it might be helpful:

tc filter show dev br0
cat /tmp/bwdpi/qosd.conf

No problem to help out here. How do you disable FreshJR?...I just see an Uninstall option.

 

[Michel]

There were many discussions about VPNs and using a different interface than br0 or eth0, but I don't recall any about using multiple interfaces for running QoS on. I'm don't know much about networking, but I don't see how you'd run QoS over multiple interfaces, it seems like you'd have to specify one interface for outgoing traffic and one for incoming traffic that you're applying the rules to. So @Michel if you think about your network configuration from that perspective maybe it will be more clear where to apply the rules.

Thanks for the reply @solstyce
I was wondering the same thing about how would QoS work for multiple interfaces. I thought maybe someone knew about it but I will have to go another way to get my topology right.

 

[Sinner]

Yes Dave, freshJR uninstalled as it would screw everything with the new categories.

As Merlin told on another thread, now the undefined category goes to the 4 one (the one you defined in the middle). Instead of going to the bottom

im not sure I like it like that.. the two MAIN features of freshJR originally was being able to put "undefined" where you needed it and the correction the garenteed bandwidth and ceiling values. these are the two most important features.

The reason is many many thing are still undefined some are gaming voice apps like teamspeak discord etc, other are MOST pc online games which you may want high priority not middle. the other major improvement was the port 80/443 web gaming console downloads into downloads instead of web browsing.. the rest was just kewl fluff features

I don't understand why they don't put "undefined" into others from factory.. major problem soved

 

[Jack Yaz]

I've stumbled on a new (I think) speedtest site from Cloudflare that seems to also do a decent job breaking down latency. https://speed.cloudflare.com/

 

[joe scian]

I've stumbled on a new (I think) speedtest site from Cloudflare that seems to also do a decent job breaking down latency. https://speed.cloudflare.com/

seems to be way inaccurate on MY download speed - reporting 56Mbps download when I have a 50/20 plan. Most other sites are correct showing 45-46 Mbps download. Still think this is the best of the lot. https://www.nperf.com/en/ and the nPerf mobile app is even better.

 

[AntonK]

I've stumbled on a new (I think) speedtest site from Cloudflare that seems to also do a decent job breaking down latency. https://speed.cloudflare.com/

It seems to be doing a good job with my Comcast 250/6 connection. Downloads showing roughly a 10-15% drop from advertised maximums, which is what I have my QOS speeds set to.

 

[dave14305]

Has anyone ever set a custom iptables rule or appdb rule using a wildcard Mark (e.g. 13****)? Seems to result in a syntax error because it doesn’t convert the asterisks to zeros before issuing the final commands. I will try to fix this if I’m right, so I just want to confirm if anyone else has used this feature?

You can see the problem if you then run the script with -debug2, plus there will be iptables errors in your syslog from the firewall-restart.

 

[andresmorago]

seems to be way inaccurate on MY download speed - reporting 56Mbps download when I have a 50/20 plan. Most other sites are correct showing 45-46 Mbps download. Still think this is the best of the lot. https://www.nperf.com/en/ and the nPerf mobile app is even better.

same for me. getting 700 on a 150 download link

 

[Marin]

same for me. getting 700 on a 150 download link

It is fairly accurate for me so far...


Sent from my iPhone using Tapatalk

 

[CriticJay]

I've stumbled on a new (I think) speedtest site from Cloudflare that seems to also do a decent job breaking down latency. https://speed.cloudflare.com/

seems to be way inaccurate on MY download speed - reporting 56Mbps download when I have a 50/20 plan. Most other sites are correct showing 45-46 Mbps download. Still think this is the best of the lot. https://www.nperf.com/en/ and the nPerf mobile app is even better.

Thanks! The more speed tests, the merrier!

I tried the Cloudflare one. It's quite nice, except it doesn't test ping during download (which provides bufferbloat information).

The Netflix speed test shows ping during download, which I use to determine bufferbloat: https://www.fast.com

 

[maxbraketorque]

I've stumbled on a new (I think) speedtest site from Cloudflare that seems to also do a decent job breaking down latency. https://speed.cloudflare.com/

The best measure of latency I've found is to run a continuous ping to google.com while performing a speed test with any of the available speed test sites. My favorites are the speedtest.net application and the fast.com website.

 

[Deleted member 62525]

I have been using Adoptive QoS and FreshJ for a long time however with the laster firmware I decided to switch to Traditional QoS with fq_codel and give it a try. I have been happy so far after more than a month. I have 300/15 MB internet and no gaming app. So between streaming Netflix, Skype, Facetime, Plex and remote streaming/browsing I have been very happy with Traditional. Customizing the rules is easy and for me this is more than adequate. Very little latency among all WiFi clients. Also, memory usage has gone down by 10%. With Adoptive I would always run 90-92% but now its at 78-80%. It actually does a decent job between direct and VPN clients.

 

[LeandroBR]

There were many discussions about VPNs and using a different interface than br0 or eth0, but I don't recall any about using multiple interfaces for running QoS on. I'm don't know much about networking, but I don't see how you'd run QoS over multiple interfaces, it seems like you'd have to specify one interface for outgoing traffic and one for incoming traffic that you're applying the rules to. So @Michel if you think about your network configuration from that perspective maybe it will be more clear where to apply the rules.

As I know, It’s possible. In this case, I think using “ifb0” interface could solve this. On this way it’s possible to control the internet traffic in eth0 to multiple internal interfaces.

It’s necessary to create a mirror from eth0 ingress to ifb0 egress. Then, control all download traffic in ifb0 and upload on eth0.

But, I have no idea if this is working in asus software. I tried, I can mirror the interface, but do not control the traffic, I don’t know why and I didn’t spend too much time on this.

 

[cenizas]

I have been using Adoptive QoS and FreshJ for a long time however with the laster firmware I decided to switch to Traditional QoS with fq_codel and give it a try. I have been happy so far after more than a month. I have 300/15 MB internet and no gaming app. So between streaming Netflix, Skype, Facetime, Plex and remote streaming/browsing I have been very happy with Traditional. Customizing the rules is easy and for me this is more than adequate. Very little latency among all WiFi clients. Also, memory usage has gone down by 10%. With Adoptive I would always run 90-92% but now its at 78-80%. It actually does a decent job between direct and VPN clients.

Hi. How are you? with the traditional qos did not lose the hardware acceleration? losing wifi speed?

 

[Deleted member 62525]

Hi. How are you? with the traditional qos did not lose the hardware acceleration? losing wifi speed?

Nope. Not at all.

 

[cenizas]

Nope. Not at all.

I did the test there, it doesn't work and my speed doesn't go up from 170 (I'm 300/15)

HW acceleration Disabled - incompatible with: QoS

 

[Bensam123]

Ridiculous this thread has so much traction after so long, rightfully so as most QoS isn't very well made even in modern routers.

So, I've been attempting to prioritize one specific machine, but when I setup a rule for it (192.168.0.10/32), it results in the router eventually becoming very unresponsive even though the rule looks like it's being applied properly. It also seems to cause a lot of oscillating latency on the computer in question when the network is under load from other clients. The router will sit at 50-90% CPU utilization and download rates suffer as well even on the highly prioritized computer.

Nothing on the router is changed and removing the rule with a reset, everything returns to normal.

 

[QuikSilver]

Ridiculous this thread has so much traction after so long, rightfully so as most QoS isn't very well made even in modern routers.

So, I've been attempting to prioritize one specific machine, but when I setup a rule for it (192.168.0.10/32), it results in the router eventually becoming very unresponsive even though the rule looks like it's being applied properly. It also seems to cause a lot of oscillating latency on the computer in question when the network is under load from other clients. The router will sit at 50-90% CPU utilization and download rates suffer as well even on the highly prioritized computer.

Nothing on the router is changed and removing the rule with a reset, everything returns to normal.

Show some screenshots of how everything is configured.