[Bloat] German Router maker implements and demonstrates ingress shaping

[Sebastian Moeller]

Dear List,


in the spirit of some things might take a while, german router maker AVM just published a Video demonstrating the new ingress shaping feature of the firmware that is described as offering some sort of per-internal-host fairness (video is in German, sorry):

https://www.youtube.com/watch?v=4VCQnzHt5B0

The demonstration is't 100% perfect:

1) Demo without Ingress-Shaping: 
Sync ~97: -> 97 * 64/65 * ((1500-20-20)/(1500+26)) = 91.3769533219*
Speedtest: 85
FTP: 91-85 = 6

2) Demo with Ingress Shaping:
Sync: ~100 * 64/65 * ((1500-20-20)/(1500+26)) = 94.2030446618* (the sync improved by 3 Mbps in between the two measurements**)
Speedtest: ~60
FPT: 94-60 = 34


So the expected share would be ~94/2 = 47 Mbps. But realistically going from a 6/85 = 0.071 ratio to a 34/60 = 0.567 ratio is likely to improve things for end users more than improving from 0.567 to closer to 1 would. I have no clue how they implement this, and I also do not know how a tricked out cake with "ingress dual-dsthost rate 100 overhead 26" would perform in this test (as cake with ingress will not necessarily equalize)

Also they claim they can to set their shaper rate to > 99% of the true bottleneck rate which according to my/our experience seems quite optimistic.



*) I assume standard VDSL2/PTM here but excluded the likely PPPoE Overhead simply becsause I am not 100% sure their ISP uses that
**) Ingress shaping requires to disable the hardware offloading and so the demonstration modem-router had to be rebooted in between the two tests and hence a different sync value