QoS setup for home network

[Acps]

Trying to find the best solution to QoS my slow DSL internet at 12Mbps/1Mbps. I have messed with the bandwidth monitor and the QoS discipline/type. My household is 6 people, and prolly 30+ smart devices connected to the wifi.

Is Asus's QoS able to work well if setup correctly, or is DD-WRT a better solution? My router is RT-AC68R and it sits behind an ATT modem.

Thanks in advance for the help!

Acps

 

[CooCooCaChoo]

Oh man, 30+ devices on a 12mbs pipe... How many of those devices are actually utilizing the pipe at the same time though?

I would look into FreshJR's QoS rule set to help out. They work quite well.

If that doesn't work, might be time to get a speed upgrade. How long has it been since you looked at the speed tiers that you bought into?

I'm with Comcast and they changed the speed tiers about a year ago. If I hadn't checked it, I would still be paying ~$90 for 150/10 instead of $95 for 250/10.

 

[Acps]

My current internet is ATT ADSL 18/1.5 Mbps, in reality its more like 16/0.9 Mbps. For only being 1-2 blocks from the node I have no idea why that is the max speed they can push to me.

My current network gear is an AT&T U-verse Pace 5268AC 192.168.1.254/255.255.255.0, and behind that is my Asus RT-AC68U 10.0.0.1.255.255.255.0. My 5268AC is currently setup as allowing DMZplus mode to my RT-AC68U with an external ip being forwarded as 104.53.24.###. Current firmware on my Asus is 384.4_2.

I believe I kinda figured out exactly my issue, and its called "bufferbloat". Particulary on the upload link. Tested by doing a continuous ping on 8.8.8.8 which on average was 25ms, then running a test using "https://www.megapath.com/speedtestplus/". During the download test its jump to an avg of about 55ms, during upload it fluctuates wildy between 1000-5000ms.

Im not sure if its my QoS settings being awful, or my ATT DSL internet\router being even worse. I thought I installed FreshJR's QoS scripts correctly, I didnt do any advanced settings tweaking, just whatever was default. Ive been trying different settings over the last few weeks, often times making it worse and having to disable QoS to fix connectivity issues.

Im going to try and do factory resets on both routers and start from scratch to see if that helps at all. On another note I do have another router, its a TP-Link WDR3500, currently with DD-WRT installed on it.

Sadly I have 0 alternatives for internet, I do not have cable/fiber as an option.

Lastly, not sure if this is an issue, but is setting my upload bandwidth to 0.8 causing issues? Anything below 1 isnt even allowed if your using the Asus App, but you can in the webgui.

If anyone else has some advice or can point me in a direction to get started I would greatly appreciate it. Thanks in advance!

~~Acps

 

[FreshJR]

Did you setup proper manual bandwidth limits ?

1) Use manual bandwidth with limits set to 85-95% of your speedtest results.

Automatic bandwidth prioritizes traffic but will NOT reduce bufferbloat.
Bufferbloat is what makes your internet unresponsive when it is under load.
Bufferbloat is the main reason of implementing QOS. If you bandwidth is not
stable, the "automatic" setting of traffic prioritization is better then nothing at all.

Also fq-codel > sfq

 

[Acps]

Did you setup proper manual bandwidth limits ?



Also fq-codel > sfq

I did a factory reset on both my ATT and Asus routers, only changing settings to setup DHCP, Wifi and QoS. With NO FreshJR scripts intalled.

I had sometime to play with it today and it wasn't pretty. I was setting my manual bandwidth limits to about 85% of what my speedtest gets. Which is about 12/0.7 Mbps. With adaptive mode enabled something is way off, because the download gets throttled to 0.5 Mbps and the upload isnt being throttled at all. It didnt matter what I set the manual bandwidth too even at 50% it was the same result. Here is a copy of my NetAnalyzr: http://n1.netalyzr.icsi.berkeley.edu/summary/id=369839a0-6163-ce99bde9-84ab-4b3e-8100

If I change the settings to bandwith limiter, I can workaround the QoS by individually setting the down/upload of every single device, and it seems to be obeying the set speed limit of about 75% 12/0.8 Mbps. NetAnalyzr: http://n1.netalyzr.icsi.berkeley.edu/summary/id=369839a0-6163-fa1821b5-d236-4dfe-8f1d
So I will try assigning all my devices to the 75%, hopefully this is a workaround. Normally it is only 1 device that is using all the upload bandwith and causing a bottleneck, so if they are all limited, we should not see any bufferbloating.

Running my ping/speedtest again to check for bufferbloating, it seems to do well, there is still some varying ping times, but overall they are only about ± 20ms.

Overall this some progress and is better than nothing, but its extremely inefficient at utilizing my piss poor DSL internet bandwidth. Any suggestions or feedback would be great to try and improve this. Thanks again for the help!

##Update##

After running a few more ping/speedtests with the bandwidth limiter settings, I have started noticing packets being dropped and getting "Ping Time Outs", its usally like 1-3 timeouts, always during the upload portion of the test.

##Update2##

On a side note, I have noticed some connective issues with accessing my ATT router (192.168.1.254) via my Asus router (10.0.0.1). It has become almost impossible to access it via the web browser, often pages being unresponsive and sometimes just unable to load at all.

Also ran a Netanalyzr from ATT Router only, no Asus, via Lan. http://n2.netalyzr.icsi.berkeley.edu/summary/id=36a470be-7675-e026a9e2-748c-48fb-bccc

~~Acps

 

[FreshJR]

Which is about 12/0.7 Mbps. With adaptive mode enabled something is way off, because the download gets throttled to 0.5 Mbps and the upload isnt being throttled at all.

Are you sure you are not inputting the numbers backwards.

The input boxes go Upload then Download when reading from top to bottom on the QOS page.

This isn't the first time I've seen reversed inputs. (Probably wont be the last).

For bufferbloat numbers on speedtest websites, FreshJR installed/uninstalled won't make a difference. The script makes changes to produce better bandwidth allocation between different types of traffic.

When all traffic is the same source, such as a bufferbloat speedtest, you won't see any changes.

 

[modemboy]

IMHO put your AT&T device into bridged mode and bypass it's routing entirely. https://wiki.sonic.net/wiki/Pace_5268AC#Bridged_mode

DSL reports speedtest has a nice buffer bloat measure built into it.