Bufferbloat and QOS Asus AC66 Merlin

[mogulman]

I'm running .58 Merlin firmware on an AC66.
ISP Comcast, 180/24 are my measured speeds. Never noticed issues with bandwidth, VoIP, etc.

Mostly use my network for VoIP, Home office vpn, video streaming, gaming pc and ps4, web, email. not really using torrents. Also other than me, the rest of my family are pretty light users, so most of the time I don't use a bunch of things that use a lot of bandwidth at the same time. When I'm on the PS4 usually the only other thing using bandwidth is my wife browsing the web on her android tablet.

So I had QOS turned off.

Recently I ran a DSL reports speed test and Bufferbloat was rated as a C. Turning on QOS makes this better, but is it really needed for a user like me. QOS limits bandwidth so I want to make sure it is really helping with something and not just making me feel better about the test.

 

[sfx2000]

QoS doesn't limit your bandwidth - just set the max values * 95 percent, and you'll be good to go

 

[mogulman]

QOS does limit bandwidth for individual tasks. If I turn it on, speedtests show my bandwidth for downloading goes down.

The bigger question that I'm wondering is, will QOS really help with anything?

 

[RMerlin]

QOS does limit bandwidth for individual tasks. If I turn it on, speedtests show my bandwidth for downloading goes down.

The bigger question that I'm wondering is, will QOS really help with anything?

You will have to chose: full speed, or lower latency at the expense of a mere 2-5% of your max speed.

There's no magic bullet there. You will also have to try it for yourself, as the results are highly dependent on your ISP and your unique usage patterns.

 

[ChatmanR]

So, if I have a 95/95 connection; my settings should be 90/90?

 

[mogulman]

On the ACC66, there is no adaptive QOS right?

If I enable Tradional QOS, then NAT Acceleration is turned off. I'm assuming this is by design. When I disable NAT acceleration, this lowers my throughput also.

I keep seeing things in certain Merlin firmware (related to Asus code) that QOS is broken in certain routers. Does that pertain to the AC66? I'm running 380.59 now.

 

[ChatmanR]

On the ACC66, there is no adaptive QOS right?

If I enable Tradional QOS, then NAT Acceleration is turned off. I'm assuming this is by design. When I disable NAT acceleration, this lowers my throughput also.

I keep seeing things in certain Merlin firmware (related to Asus code) that QOS is broken in certain routers. Does that pertain to the AC66? I'm running 380.59 now.

mogulman, I'm using the same model. As far as I know, the AC66 doesn't have Adaptive QOS. I'm going to try to set mine up using the guide from linksysinfo http://www.linksysinfo.org/index.php?threads/qos-tutorial.68795/#post-231133

 

[mogulman]

These are my results with QoS off, NAT acceleration on:

 

[mogulman]

When I enable QoS (Nat acc off), QOS set for 185 down, and 22 up.

Nat Acc doesn't seem to make a big difference. Tried the first test with Nat Acc off also.

So my speed is way down for download with QoS turned on. Why so much? It's consistantly in the 99Mbs area rather than 180 with QoS enabled.

 

[ChatmanR]

When I enable QoS (Nat acc off), QOS set for 185 down, and 22 up.

Nat Acc doesn't seem to make a big difference. Tried the first test with Nat Acc off also.

So my speed is way down for download with QoS turned on. Why so much? It's consistantly in the 99Mbs area rather than 180 with QoS enabled.

Why did you choose 185 down, and 22 up? Shouldn't be closer to 150/20?

 

[mogulman]

Why did you choose 185 down, and 22 up? Shouldn't be closer to 150/20?

Initially I tried 170, then 180 and then 185. It didn't seem to make a difference for the download speed.

Anyway..all 3 settings had the same thing for download. Limiting my download to 97-99Mbs.

I tried 21-22 Mbs for upload also.

 

[sfx2000]

You're going off one benchmark - try some others...

 

[mogulman]

Ran speedtest at 3 different sites. Xfinity, Ookla, DSL Reports. QoS set for 170/21. All report 98-99Mbs download with QoS on. With QoS off, 180Mbs.

Any ideas? I tried .58 and .59 firmware with same results. I also tried Restort Default Factory settings after installing firmware.

 

[sfx2000]

iperf3 -c iperf.he.net -i1 -Z

this will give you uplink... add the -R switch for downlink..

search for iperf3 - you can download iperf3 binaries for Windows - https://iperf.fr/iperf-download.php

 

[sfx2000]

And make sure you're wired, and not on the WLAN, you want the real performance - iperf3 will do that for you...

 

[mogulman]

That's a good tool, but I only get about 33Mb download with it, even without QoS, so doesn't really help me. I don't really have the resources to set up a local iperf or jperf server.

Anyone have any idea why QoS would be limiting download like that?

I am wired btw.

 

[sfx2000]

I'm running .58 Merlin firmware on an AC66.
ISP Comcast, 180/24 are my measured speeds. Never noticed issues with bandwidth, VoIP, etc.

Mostly use my network for VoIP, Home office vpn, video streaming, gaming pc and ps4, web, email. not really using torrents. Also other than me, the rest of my family are pretty light users, so most of the time I don't use a bunch of things that use a lot of bandwidth at the same time. When I'm on the PS4 usually the only other thing using bandwidth is my wife browsing the web on her android tablet.

So I had QOS turned off.

Recently I ran a DSL reports speed test and Bufferbloat was rated as a C. Turning on QOS makes this better, but is it really needed for a user like me. QOS limits bandwidth so I want to make sure it is really helping with something and not just making me feel better about the test.

Then maybe don't worry so much about bufferbloat - at 100Mbs on the WAN side, it's not really an issue, revert back to what you had initially and all is good...

 

[gffmac]

Without QOS I get this http://www.dslreports.com/speedtest/4318982 my bufferbloat (C) doesn't improve even if I quarter my upload speed. Must be my ISP.

 

[Rich Brown]

You're going off one benchmark - try some others...

Yes. DSLReports.com provides an easy test, but you can get a more quantitative test by starting a ping to a remote host, then using any of the speed test sites to see if your ping times jump up.

That will give you a sense of whether/how much heavy traffic affects latency/lag/ping times. It also allows you to make a change to the router config, and see the results.

The full procedure is at Quick Test for Bufferbloat ...

 

[mogulman]

Yeah. I tried that also...

Goes from 10ms ping to like 250ms for downloads.... 125ms for uploads...
I think if I could get QOS working without losing half my download capacity I'd stick with it.