DSLReports Speed Test Now Measuring Buffer Bloat

[DaBiGoReO]

For anyone following this thread with the RT-N66x and John's fork - I had a heck of a time getting this dialed in. Before any additional changes and QoS already on, I got Bufferbloat:C and Quality:A. The best I could get after A LOT of tweaking and retesting was Bufferbloat:A and Quality:B. I'm on Fios with 50/50.

My final settings were:
Upstream BW: 45Mbps
Downstream BW: 41 Mbps

My bandwidth settings for QoS seem high but they actually test much higher. I also had to set download caps in the User-Defined Priorities section (65%). This control does seem to make a difference based on my testing even though Asus says it doesn't. What is odd about this is even though my max downstream is set to 41Mbps in QoS, I still get close to 60Mbps when testing. The affect I'm seeing is this helps with bufferbloat control while sacrificing some quality. YMMV as they say on the web. Hope this helps anyone playing with this on the RT-N66x.

 

[tsunami2311]

I do need it cause i have 20 wifi device connect to the router and i dont want any of them monopilizing the bandwidith.

And i dont want to put the QOS to 95mbit, cause I want the Wired connection to have max bandwidth possible, if I do that they dont get it

Unless there another way to make it so Wired has max speed possible, while still limiting wifi and proitizing gameports with QOS off

 

[Nullity]

I do need it cause i have 20 wifi device connect to the router and i dont want any of them monopilizing the bandwidith.

And i dont want to put the QOS to 95mbit, cause I want the Wired connection to have max bandwidth possible, if I do that they dont get it

Unless there another way to make it so Wired has max speed possible, while still limiting wifi and proitizing gameports with QOS off

What is "it"? Adaptive QoS?

Unless it separates each device/host into an individual queue (a modified Fair Queueing), any device could saturate the link.

To separate WiFi and wired clients, many people use VLANs or separate DHCP pools, among a number of other methods. With either choice, allowing different bandwidths for wired and WiFi traffic is easy.

 

[Mears148]

For anyone following this thread with the RT-N66x and John's fork - I had a heck of a time getting this dialed in. Before any additional changes and QoS already on, I got Bufferbloat:C and Quality:A. The best I could get after A LOT of tweaking and retesting was Bufferbloat:A and Quality:B. I'm on Fios with 50/50.

My final settings were:
Upstream BW: 45Mbps
Downstream BW: 41 Mbps

My bandwidth settings for QoS seem high but they actually test much higher. I also had to set download caps in the User-Defined Priorities section (65%). This control does seem to make a difference based on my testing even though Asus says it doesn't. What is odd about this is even though my max downstream is set to 41Mbps in QoS, I still get close to 60Mbps when testing. The affect I'm seeing is this helps with bufferbloat control while sacrificing some quality. YMMV as they say on the web. Hope this helps anyone playing with this on the RT-N66x.

I actually posted about the Adaptive QoS settings not really limiting bandwidth just yesterday in a separate thread. I wondered about this too and whether this was the intended behavior.

 

[DaBiGoReO]

I actually posted about the Adaptive QoS settings not really limiting bandwidth just yesterday in a separate thread. I wondered about this too and whether this was the intended behavior.

To my knowledge, the RT-N66x does not have Adaptive QoS but only traditional.

 

[tsunami2311]

AC66U dont support adative it either QOS or Bandwidth limiter

 

[Nullity]

I think Adaptive QoS & limiters are featured exclusively on ARM devices.

 

[Badlandz]

I also had to set download caps in the User-Defined Priorities section (65%). This control does seem to make a difference based on my testing even though Asus says it doesn't.

I can confirm this. In my testing I could not get better than a 'C' grade until I also tweaked the download cap as well. There is a sweet spot as higher settings seem to do nothing. It was the only way I was able to get an 'A' grade.
- I'm on John's fork V15E5, AC66U.

 

[Badlandz]

Setting the download cap in qos gives me an 'A' in bufferbloat but the quality and speed take a nose dive and my ingress bandwidth is 60% of normal.

Tried using the tc command in user scripts, but can't seem to get a seperate flow rate for upload and download.

In the end I gave up and used the gui qos and an user script to del the ingress qdisc and make a new one. This approach gave both an 'A' in bufferbloat and an 'A' in quality (which I assume is the benchmark used for dropped packets).

 

[Nullity]

I thought quality was latency fluctuation/jitter, but I am not sure. Dropped packets makes more sense.

 

[bayern1975]

this is mine test from dslreports....missing top speed parameter but do not know why.....

 

[Roy2001]

Adaptive QoS + my down/up rates set to roughly 90% of the maximum rate (I have 30/10 cable):

No need for special arcane algorithms.

This works for me, and BufferBloart is improved from F to A. I have Comcast 100Mb/5Mb, actual speed is 120Mb/6Mb and I set to 110Mb/5.4Mb. Thanks

 

[sfx2000]

It's easy to trick perhaps... but a little pfSense magic with perhaps an overpowered router...

150/10 cable...

 

[hispanico957]

With AC-68U and 360.2


Hispa

 

[david279]

Like the speedtest.....

 

[jdpg2]

Only way to get A's across the board was to set my download speed < 10% less then the max

Adaptive QoS Manual settings : 200 down , 11 up

Adaptive QoS Automatic ( Comcast says 250 down, 12 up)

 

[sfx2000]

Interesting - look at the number of streams in the different "badges" - wonder what makes that change...

Some are 24/12, some 8/6, some 32/6 - weird...

 

[RMerlin]

Interesting - look at the number of streams in the different "badges" - wonder what makes that change...

Some are 24/12, some 8/6, some 32/6 - weird...

Faster links use more streams to better saturate the link.

 

[calzor suzay]

Always remember to reboot your cable modem once in a while...

I was running tests and getting this, I should be on 200/10

Couldn't work out why but was pretty consistent so decided to pull the plug on it for a bit and try again
I then got this without any QOS

So I set it on Adaptive QOS at 200/10 and got this

 

[doczenith1]

The number of streams maxes out at 32 for their "1 gig" test.

When I run the 1+ gig test it only uses 24 streams and bufferbloat starts rearing it's ugly head.