Cell Relay Archive

[albert.e.manfredi@boeing.com]

In article <86btmwiljy.fsf@ironbridgenetworks.com>,
  James Carlson <carlson@ironbridgenetworks.com> wrote:

> Lack of service is just lack of service.  It doesn't much matter to
> the user if the reason that the service doesn't work is because the
> intermediate links are overrun, or if the server is dead, or if the
> local applications are hung.
>
> The price for that "busy signal" is that every intermediate node is
> keeping state -- it knows about the existence of this one small
> peering.  That might make some sense in a small network, where
> services are under the administrative control of a single domain, but
> it makes little sense in a public network backbone, where the laws of
> large numbers apply.

Yes, but the law of large numbers does not say that certain sessions will not
encounter persistent congestion. That's the issue here. With IP, even with
some attempt at queue priorities, persistent problems can and will occur
almost for sure, even if not on a majority of routes. They will occur on the
most used routes.

> (Note this domain problem is large.  What if a "high priority"
> customer connects to a "low quality" of peer?  Which priority applies?
> What if the connection is made in reverse?  These problems all run
> into the reservation issue -- if you make reservations, then you
> effectively channelize and you maintain state storage.)

That's why ATM signaling is not based on priorities as such.

Bert
manfredi@arl.bna.boeing.com


-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own