[Nsi-wg] 3rd Party connection requests

John Vollbrecht jrv at internet2.edu
Tue Feb 9 10:45:32 CST 2010


Hi Jerry -- thanks for this work -  I have not been able to digest it completely.  I am working to try to make some distinctions so that your work can be incorporated in recommendation in a  way that is distinct from other parts.  I hope to have this in the next few days.  

Meanwhile we can talk about this on the call tomorrow -

John

On Feb 7, 2010, at 8:31 PM, Jerry Sobieski wrote:

> I wish to pose two questions for discussion regarding Connection Service 
> request handling:
> 
> 1st:  Can we reserve resources backwards? 
> 
> 2nd:  How should third party requests be handled?
> 
> Backward Reservations-
> 
> This is simply doing a path finding process backwards - from destination 
> to source.   This does not (must not) change the direction of the 
> connection that is established in the transport plane, it simply 
> constructs the viable path tree from the destination towards the 
> source.  THis should still develop a valid connection path, but it may 
> not be the same path selected by a forward pathfinding process...
> 
> My recommendation:  It should not matter whether a path is discovered 
> from Source to Destination or vice versa.  And it is up to the local 
> domain as to decide how to decompose a connection request and perform 
> the path finding/resource allocation within its domain.   Any 
> intermediate hops specified in a PO must be honored.
> 
> 3rd Party Requests
> 
> A "3rd Party Request" is a request receive by a local NSA that specifies 
> end-points that are not located within the local NSA's domain.  I.e. 
> there is no basis for assuming that the [shortest/best] connection path 
> will transit the local domain.  So the question is: should the local NSA 
> simply forward the request towards the Source (or Destination) ? or can 
> the local NSA insert a local intermediate STP (a loose hop) in the path 
> object, thus forcing the connection to transit the local domain?  
> 
> Since this has to do with how we handle a request - not particularly 
> pathfinding per se, I think we need to think about this to decide if we 
> need to make some firm declaration about this.  
> 
> One potentially questionable result of this issue is that an 
> intermediate NSA who handles a request could insert a local hop into the 
> Path Object, thus forcng the connection thru their domain.   If this 
> sounds ominous, it is.   So we should consider what the desired action 
> should be and how to be certain it is followed.
> 
> Thoughts?
> Jerry
> _______________________________________________
> nsi-wg mailing list
> nsi-wg at ogf.org
> http://www.ogf.org/mailman/listinfo/nsi-wg



More information about the nsi-wg mailing list