Traditional QoS performance

[noxide]

Above is without QoS

Above with QoS with those settings.

It looks like the upload is sorta working but what's going on with the download? Adaptive QoS works, but it means I must accept the privacy thing so I'd rather not.

CT8 mesh setup, tested with wired connected directly to the main node (router).

 

[bbunge]

View attachment 46287

Above is without QoS

View attachment 46291

View attachment 46292

Above with QoS with those settings.

It looks like the upload is sorta working but what's going on with the download? Adaptive QoS works, but it means I must accept the privacy thing so I'd rather not.

CT8 mesh setup, tested with wired connected directly to the main node (router).

With a GB download why are you using QOS at all? The Adaptive works very well for me with 100/100 FIOS...

 

[noxide]

With a GB download why are you using QOS at all? The Adaptive works very well for me with 100/100 FIOS...

Bufferbloat.

 

[hancox]

Believe traditional turns off hardware acceleration, hence the cap at ~250-300 Mbps.

ASUS would suggest to use Adaptive, which doesn't have this issue, but it requires TrendMicro enabling, but it's also outright broken on at least the AX86U, for example, so YMMV.

 

[bbunge]

Believe traditional turns off hardware acceleration, hence the cap at ~250-300 Mbps.

ASUS would suggest to use Adaptive, which doesn't have this issue, but it requires TrendMicro enabling, but it's also outright broken on at least the AX86U, for example, so YMMV.

Nope, as I stated the Adaptive QOS works for me on an Ax86U!

Still, with GB bandwidth why worry about bufferbloat?

 

[hancox]

Nope, as I stated the Adaptive QOS works for me on an Ax86U!

Still, with GB bandwidth why worry about bufferbloat?

It only works because you have 100/100, so not near the point where hardware acceleration would come into play.

Anyone with above ~200Mb/s will see issues.

 

[noxide]

Nope, as I stated the Adaptive QOS works for me on an Ax86U!

Still, with GB bandwidth why worry about bufferbloat?

Latency suffers with Virgin Media. My upload is crap too.

 

[Tech9]

Your only option is Adaptive QoS, if you need QoS at all. Don't measure bufferbloat with online tools. Most of the time in normal use you have no bufferbloat on Gigabit ISP. Traditional QoS is NAT acceleration incompatible and depending on your router model WAN-LAN throughput will be cut down to 200-400Mbps range.

 

[noxide]

Your only option is Adaptive QoS, if you need QoS at all. Don't measure bufferbloat with online tools. Most of the time in normal use you have no bufferbloat on Gigabit ISP. Traditional QoS is NAT acceleration incompatible and depending on your router model WAN-LAN throughput will be cut down to 200-400Mbps range.

I wasn't aware NAT acceleration is disabled. Thanks.

 

[BreakingDad]

My life has improved since the removal of qos

 

[noxide]

My life has improved since the removal of qos

Useful info.

 

[BreakingDad]

Useful info.

I have 550/38 , 1 kid has flown the nest, 2 still at home. Very litle spiking these days. We "try" not to download when someone gaming.

Spent so long trying to get QOS to work, in the end things seems better without it.

 

[noxide]

I have 550/38 , 1 kid has flown the nest, 2 still at home. Very litle spiking these days. We "try" not to download when someone gaming.

Spent so long trying to get QOS to work, in the end things seems better without it.

I reverted back to my old configuration with SQM cake on a router and using the mesh just for AP. Was trying to simplify things.

 

[Tech9]

What router is running Cake with Gigabit speeds?

 

[noxide]

What router is running Cake with Gigabit speeds?

WRT32X

 

[Tech9]

Interesting. Marvell Armada CPU shows some muscle.

 

[RMerlin]

Interesting. Marvell Armada CPU shows some muscle.

Or NAT acceleration is caching the flows, and therefore Cake doesn`t do anything at all.

 

[noxide]

Or NAT acceleration is caching the flows, and therefore Cake doesn`t do anything at all.

How would I test for that? SQM is definitely working as I can cap my speeds and it sorts out a lot of the bufferbloat according to dslreports

 

[Tech9]

Test with and without QoS and see if there is any real use difference. Bufferbloat test sites create the issue by saturating your ISP line and then you fight a non existing problem 99% of the time by hurting your ISP connection 100% of the time. I'm also in doubt this ARMv7 cores based CPU in WRT32X can do true packet processing with Gigabit speeds. Some sort of NAT acceleration hack is in play.

 

[noxide]

Test with and without QoS and see if there is any real use difference. Bufferbloat test sites create the issue by saturating your ISP line and then you fight a non existing problem 99% of the time by hurting your ISP connection 100% of the time. I'm also in doubt this ARMv7 cores based CPU in WRT32X can do true packet processing with Gigabit speeds. Some sort of NAT acceleration hack is in play.

Dslreports hasn't saturated my line for a while. Regardless of my setup I mostly see 400 or 500Mb/s
I've ran this WRT32X setup for a long time. I've been wanting to move over to a XT8 or above for AX and to make use of the 2.5GbE on my modem. I was trying out the CT8 in router mode and PiHole. It did work fine but according to a broadband monitor via ThinkBroadband over a two day period was smoother with my WRT32X setup. I assumed it was the SQM.

With the WRT32X I've been testing the NAT acceleration options under Firewall and packet steering. They've been off up until now.