The MPLS WG Archive

[Chris Cooper <chris.cooper@rl.ac.uk>]

Fred Baker wrote:
> 
> At 03:00 PM 9/26/00 -0400, Martin Picard wrote:
> >   Service Providers tend to say that their backbone will never be
> >   congested and if it ever gets there, then, bandwidth will be increased
> >   and therefore no Congestion Management or Congestion Avoidance
> >   mechanisms are necessary.
> 
> I know of a number of service providers, including BT etc, who are
> deploying services using QoS technologies. Whether or not the specific ones
> I am talking with are using MPLS in the same application, I can't say; I
> don't know, and I think some are and some aren't.
> 
> The above statement is indeed what most of the SPs say, but it misses a
> rather important fact. They are engineering their core to have enough
> bandwidth that they experience a low drop rate. That means that QoS
> perceived by their customers is not limited by their core. But it remains
> limited by the link between the edge network and the SP, which is supplied
> by the customer, and the delay and drop rates at the service provider end
> of that link are invisible to him. As often as not, it is specifically this
> link that will benefit from QoS technologies.

Yes, exactly.  I also have encountered this sort of assertion recently from
SPs.  But the basic issues seem not to have changed: see for example the
summary of the infinite bandwidth argument provided in RFC1633 by Bob
Braden, Dave Clark, and Scott Shenker in 1994.

Chris
-------------------------------------------------------------------
Chris Cooper			Email:	chris.cooper@rl.ac.uk
Rutherford Appleton Laboratory	Tel:	+44 (0)1235 446211
Chilton 			Fax:	+44 (0)1235 445597 (NB.NEW)
Didcot,  Oxon  OX11 0QX,  UK