The MPLS WG Archive

["Philip Matthews" <philipma@nortelnetworks.com>]

MPLS wrote:
> 
> Hi
> 
> I am wondering if in a network we are running CR-LDP TE (for traffic
> having QoS requirements as well as BE traffic), how do we account for the
> bandwidth used by best effort traffic while doing QoS path computation for
> flows having QoS requirements.
> 
>  Do we consider bandwidth required by BE traffic as zero or we take
> average bandwidth used by BE flows into account. If we account BE LSPs
> having zero bandwidth then how accurate are the QoS path computations if
> the BE flows account for a substantial portion of the network traffic.
> 
> Manish

As you yourself implied, it is clearly necessary to make unlabelled
traffic
into account when setting up LSPs for QoS traffic.

If one assumes that QoS traffic is generally more important, then
one solution is to make the weights used by the IGP for route
computations
a function of the unreserved bandwidth on a link. As the bandwidth
reserved for QoS LSPs increases, this will force the unlabelled traffic
onto other links when possible.

Peter Ashwood-Smith's MPLS tutorial at Networld+Interop this week has a
short
section on this topic. You might wish to attend and/or look at the
slides.

- Philip