[Cake test] Cake experiments for 386.2 Beta

[kernol]

I get by far best results with cake-qos as built into 386.2 beta1 [triple A+ and best speeds 186d and 87u]- these mods make it worse [B for quality and C for bufferbloat - speeds 156d and 74u] ... but no doubt to be expected given my more symmetric Fibre.

I assume these test are intended for asymmetric connections.
Reverted to built in {ThumbsUp}.

 

[kernol]

I think it may be because we don’t yet have the beta2 build where qos-start is a blocking script.

On point ... . Only 3 models have beta2 build - so rest of us wasting time and confusing the results - sorry

 

[rlj2]

On point ... . Only 3 models have beta2 build - so rest of us wasting time and confusing the results - sorry

Is that what my issue may be? Title says 386.2 beta, not necessarily 386.2 beta2... Im using 386.2 beta 1

 

[DanNuke86]

I'm confused by this confusion.. thought it was beta 1 this applied to?

 

[dave14305]

Is that what my issue may be? Title says 386.2 beta, not necessarily 386.2 beta2... Im using 386.2 beta 1

It’s a minor change in beta2 that may limit the effectiveness of this qos-start script for this purpose in beta1. Depends what version Merlin was running when he came up with this idea.

 

[rlj2]

It’s a minor change in beta2 that may limit the effectiveness of this qos-start script for this purpose in beta1. Depends what version Merlin was running when he came up with this idea.

It just seems like my ifb4eth0 wont go away, think someone else said that.
my /etc/cake-qos.conf has
ULIF='eth0'
DLIF='eth0'
MIF='ifb4eth0'

I dont see if the script where MIF would get deleted. Maybe its not necessary.

 

[RMerlin]

It is better to use the WAN interface for uploading and the LAN interface for downloading, because this problem does not happen.

That was also my initial impression. I decided to have this little experiment with a few users anyway because I know for instance that the Trend Micro bwdpi uses the LAN interface for its downstream. However I am aware that their implementation is quite different (they internally access with Broadcom's private APIs for some stuff), so I wouldn't be surprised if in Cake's case this might not work properly.

It’s a minor change in beta2 that may limit the effectiveness of this qos-start script for this purpose in beta1. Depends what version Merlin was running when he came up with this idea.

That`s possible, I do believe the two routers where I have done some quick tests here were both running pre-beta2 builds of some sort.

So far, I'm leaning on sticking with the original implementation. It might not be optimal performance-wise, but it's probably safer and more reliable as a whole. Things may get complicated if we handle br0 traffic, as we also have to take into accounts people using VLANs, and so on.

 

[rlj2]

Looks like this may not be a direction things are heading anyways. But I like to test. I did a factory reset on the AC86 just to make sure there wasnt a glitch in my setup, and 4 nodes. But the outcome was just the same.

 

[RMerlin]

Can't you test br0 or another interface yourself, to see if the problem I mention happens? maybe i'm wrong.

My connection is too fast, I need hardware acceleration enabled.

 

[RMerlin]

Closing down this experiment, thanks for testing folks.