For understanding purposes essentially i have 3 doubts about that technology:
1) if the weights of the flows are based on IPP and the IPP are 3 bits the maximum number of combinations i suppose then would be 2^3
So in the case we have more flows what happens? Probably is cause of that that may occur hash collision?
2) With the weighting based on IPP, the scheduler will serve the flow based on the highest IPP so the mechanism works because the weight is inversely proportional of IPP and also of the proportional share of bandwidth. But this condition is not satisfied when the weight is equal among the flows. If we have 2 flows, for example, with same weight then the scheduler will select for the tx-ring the flow with smaller packets. Is that correct? because if it's correct it should have a negative impact on the manage of some flows since the IPP discriminant settings are revoked.
3)The hold queue buffer space is proportionally shared across all flows or is just shared across the entire software queue?