Congestion Control Sliders

[Dan]

Hi Netduma

I am currently on the latest software 1.03.6h

I'm on BT infinity with the R1 connected directly to the openreach modem

Usual speeds are c.36mb download and c.9mb upload

Using dslreports for testing in chrome

Using preemptive settings

 

Example of a sets of test are below

Issue: Congestion Control (CC) for download is exaggerated

at 100% CC download 34mb

at 99% download falls to 22mb; a 35% fall

at 90% 20.4mb  40% fall

at 80% 16.0mb  53% fall

at 70% 14.4mb  58% fall

 

Issue: Congestion Control (CC) for upload is exaggerated

at 100% CC upload 8.5mb

at 99% download falls to 7.8mb; a 8% fall

at 90% 7.7mb  9% fall

at 80% 7.2mb  15% fall

at 70% 7.0mb  17% fall

 

I have gone all through the tutorials for fine tuning the R1 (e.g reset distribution, share excess, allow the R1 to configure bandwidth settings, turn off IPv6, network quiet when testing, Turbo off, Super Turbo off, deep packet off)

 

As you can see CC changes to upload are understated but CC changes to download are massively overstated. Even a 1% CC reduction knocks off 35% of total speeds for the household. It looks like the CC algorithms are all of sorts.

 

Occasionally when making bandwidth changes I get an error message that says "unable to apply a CC algorithm....."

 

Grateful for any thoughts

[RL317]

I only notice a discrepancy* with my download if I'm using the preemptive algorithm. For example I get 50/10Mb, so if I was to use 70/70% sliders I'd get something like 25/7 on a speed test. If I use reactive I get whatever I tell the sliders I want.

 

*This is by design; preemptive uses Netduma's own patented QoS algo which is more aggressive when it comes to combating congestion. This is why if you have bandwidth over a certain amount it's generally recommended that you use reactive if you care about how much bandwidth you have left to play with. For gaming I've found preemptive is best because it gives me the best stability, and reactive gives me packet loss lol

[Netduma Fraser]

As above Preemptive is more aggressive than reactive which is more conservative. You're likely to get a better connection by going with preemptive and are unlikely to notice a discernible difference with actual downloads. You can of course experiment with reactive also.

[Dan]

Thanks for your comments. In my situation preemptive is very aggressive on the download but conservative on the upload.

 

On reactive the upload ping is very erratic might be 10ms or 100ms or much worse

 

What happens when updating Bandwidth without auto configuration? Does changing the bandwidth numbers just effect the CC scale?

 

When changing bandwidth and saying Yes to auto configuration does this just make changes to pages like preemptive / reactive adding deep packet and enabling ipv6 etc or does it do other things that can't be seen on the GUI?

[RL317]

Are you using the device prioritisation flower? That makes up and down values wonky for me. Even with 100% sliders and share excess ticked, my consistent 52/10 drops to 40/9 if I play around with the flower, and it gives me packet loss even with share excess left on lol. Try hitting reset distribution under that and testing again.

 

Auto configuration doesn't really change much; it just enables or disables certain settings like IPv6, turbo mode or deep packet processing depending on what you put in, so it's good for those who aren't too network savvy and don't know which settings do what. It also chooses preemptive or reactive for you depending on your bandwidth. Just put your true bandwidth numbers in, click no for auto setup and adjust the other settings to your liking.

 

For example I find it best to use preemptive, IPv6 off in WAN and LAN, and all miscellaneous settings disabled apart from cookies enabled. Keep it nice and simple, meaning there's less for the router to do

[Dan]

Thanks IIIRL

 

Most of the time I may only have four main devices on the flower and 1-2 will be dormant. All the above testing is on a quiet network, post reset distribution and share excess ticked. Where my experience differs is that ticking share excess does exactly that; regardless of the flower the PC doing the testing gets everything (because no one else is using it even though they are in the flower). The internet analysis tool always gives me exceptional and no packet loss. Also agree with your sentiments no IPv6 and in misc only have cookies and stealth mode.