Re: delay pools and VIP

On Mon, 13 Sep 1999, David Luyer wrote:

>
> Milos Prudek wrote:
>
> > Let's say I need priority access for a single IP (let's call it VIP).
> > Whatever bandwidth that IP asks for, it will get it. If inactive, other
> > IPs will get unlimited access. Seems to me I can't set delay pools to do
> > exactly that.
> >
> > Given our capacity of 3000 B/s, I can say that VIP can have no more that
> > 2500 B/s, and other IPs can have no more than 1500 B/s. That's fine when
> > both VIP and IPs are surfing, but when VIP quits, the IPs are still
> > limited to 1500 B/s, right?
>
> Quite correct. HTTP is stateless. There's no definition of when VIP is
> surfing or not surfing to start with.

I have to disagree in this particular context.

IMO, the definition of "when VIP is surfing" is "when there is a pending
request from VIP". Technically, it should be possible to limit others'
bandwidth just for the time when VIP request is served. I understand
that there may be some edge-effects (e.g., some packets are still coming
in for low-priority IPs), but those effects will be limited to TCP
buffer sizes (which could be, theoretically at least, adjusted run-time
as well).

I am not saying it is easy to implement, just trying to make a point
that it is feasible.

Alex.
Received on Mon Sep 13 1999 - 11:03:42 MDT