Re: Lambda LSP establishment

[Yakov Rekhter <yakov@juniper.net>]

Maarten,

> Eric,
> 
> In a parallel discussion it was concluded that an IETF FA is equivalent to a
> G.805 link. And within ASON, routing adjacencies are over G.805 links... i.e.
> over an FA in IETF terms.
> 
> I believe that we are hitting here a major difference between GMPLS and ASON
> (and between GMPLS and existing management plane based connection management
> (eMPbCM)).
> 
> ASON (and eMPbCM) establish layer network X routing adjacencies between two
> layer network X switches which share (i.e. are at the two ends of) a G.805 
> layer network X link.
> 
> GMPLS establishes routing adjacencies between two equipment which share a
> physical medium (e.g. fiber).

That is certainly *a* way to establish routing adjacencies in GMPLS.
But not the *only way possible*, as GMPLS does *not* restrict
routing adjacencies to only "two equipment which share a physical
medium".
  
> This difference in my current understanding may prevent interworking 
> between an ASON/eMPbCM network and a GMPLS network.
> 
> Looking at the document introducing the FA concept
> (draft-ietf-mpls-lsp-hierarchy), I noticed that it is a MPLS document and not
> a GMPLS document... 

That is correct - it is an MPLS WG document. So what ?

> the document defines a MPLS FA.

That is correct - the document does define an MPLS FA.

> A MPLS FA is equivalent to a MPLS serial compound link (G.805); i.e. a serial

Please note that in the MPLS terminology there is no such thing
as "MPLS serial compound link".

> concatenation of MPLS links (G.805 terminology). The optical network FA (e.g.
> a VC-4 FA supporting a VC-12 link) is different as it represents a server 
> layer trail supporting a (single hop) link in the layer network of interest.
> 
> The MPLS link (G.805) between two adjacent MPLS LSRs is already a (optical
> network) FA (ON-FA). And two MPLS LSRs have no doubt a routing adjacency, but
> as indicated this adjacency is via an ON-FA... so my conclusion is that
> routing adjacencies are also over FAs.

First note that for the purpose of control there are routing
adjacencies (used for exchanging ISIS/OSPF Hellos and for Link
State flooding), and there are signaling adjacencies (used for
RSVP/CR-LDP signaling). And the presense of signaling adjacency
between a pair of LSRs doesn't imply that there is a routing
adjacency between these two LSRs.

Second node that even if two LSRs have an FA (or a link) between
then, it doesn't follow that routing adjacency *have to* be
established over this FA (or that link). In fact, there may not be
a routing adjacency between these two LSRs in the first place.

Yakov.