RE: OSPF-TE question....

["Naidu, Venkata" <Venkata.Naidu@Marconi.com>]

Venkat,

-> Is this possible:
-> 
-> 
->          +------R1--------+
->          ^   Area 0       v
->          ^                v
->         N1                N5
->        /  \             /    \
->      /     \          /       \
->    N2  area N3------N6  area   N7
->      \  2   /         \   3   /
->       \    /            \    /
->         N4                N8
-> 
-> N1-N8 are OXC nodes.
-> R1 is a router providing control plane connectivity.
-> 
-> All Na-Nb links are TDM links and provide control plane 
-> connectivity except
-> N3-N6 where control plane connectivity is provided 
-> out-of-band. All Na-Nb links
-> have TE attributes that need to be advertised.
-> 
-> Links N1-R1-N5 are not TE links.
-> 
-> If the service provider requires that N1-N4 needs to be part 
-> of area 2 and N5-N8
-> be part of area 3:
-> 1. Can N3-N6 be advertised at all?

  A link (reachability or TE) must be in one and only one
  OSPF area. If you configure N3->N6 link in Area 2 and
  N6->N3 link in area 3, then N3 and N6 won't form any
  OSPF adjacency as per RFC2328. But N3 and N6 routers 
  *MAY* (depending on the implementation) advertise their
  respective links to respective areas.

  Finally, you will end up seeing unidirectional links 
  being advertised in Area 2 (N3->N6) and in Area 3 (N6->N3)
  TE topology database.

  Unlike, OSPF SPF, there is no *rule* for bi-directionality
  test in (G)MPLS CSPF. Again, you *MAY* (depending on the
  ingress LER CSPF) successful establish a uni-directional LSP
  to N3/N6 as egress LERs (depending on area 2 or 3 CSPF).

  You may get unpredictable results in case of multi-area TE
  approaches.

-> 2. Can you couple two area (2 and 3 in this example) from a 
-> TE topology point of view but not for the regular topology?

 You may want to look at:
 
http://www.ietf.org/internet-drafts/draft-venkata-ospf-te-only-option-00.txt

-> 3. Based on answers for 2 and 3, would N3 and N6 need to 
-> establish ospf adjacency?

  see answer to question 1.

--
Venkata.