Re: [RFC/PATCH] IMQ port to 2.6

["Vladimir B. Savkin" <master@sectorb.msk.ru>]

> Ok, i think i have understood you finally;->
> The challenge is in this one direction whose characteristics can be
> described as follows:

In other direction, the goal is the same, but IMQ is not needed,
since there is only one Internet link.

> a) Incoming pipe (from internet) is smaller than outgoing pipe (to
> clients).

Yes, and artificial limit is even smaller.

> b) Desire is to have excess bwidth with max fairness to all flows
> instead of free-for-all scheme.

Yes, if you define "flow" as all traffic to one client.
Actually, I use two-level hierarchy: in every flow in above sense
each micro-flow receives a fair amount of bandwidth (approximatly,
using sfq).

> [This can only be achieved by a non-work conserving scheduler].

Yes.

> 
> Is the above correct?
> 

It seems so :)

~
:wq
                                        With best regards, 
                                           Vladimir Savkin.