Re: [mpls] working group last call on draft-ietf-mpls-spring-entropy-label

[Stewart Bryant <[e-mail]>]

On 30/04/2016 21:05, Carlos Pignataro (cpignata) wrote:
> The RLD relevant to the calculation seems to be a characteristic of 
> the path on a segment, spanning multiple nodes. 

At first sight it would seem to be min(RLD) of all of the possible paths 
that might be taken between the segment end-points. This is quite a 
large calculation, since it requires a per egress node, all paths 
calculation to any possible next segment node calculation.

However is an IPFRR repair is configured then the packet may travel on a 
path that would not normally be considered in a formal shortest path 
determination.

In a network that might configure IPFRR, I think that RLD approximates 
to min(RLD) of the network, and will of course change as the network 
evolves in both the short term (failure, re-instatement, management 
change) and the long term (reconfiguration, new links, new equipment).

The per network approximation of course massively simplifies things.

There was some work in RTGWG that Mike, Alia and I did (sorry if I 
forgot anyone else) on how to advertise such domain wide semi-constants 
since it was needed for convergence time.

- Stewart