The MPLS WG Archive

["Darek Skalecki" <dareks@nortelnetworks.com>]

I have a couple of questions/comments in association with draft-venkatachalam-interarea-mpls-te-00.txt
about LSP setup across IGP areas.

1. In section 4.4 (Routing algorithm):

   6. If the destination is in this area, calculate an intra-area route
      from the originating node to the destination that satisfies the
               ^^^^^^^^^^^^^^^^
      constraints, and splice the egress of the original LSP and the
                                                ^^^^^^^^^^^^
      ingress of the new LSP.

Shouldn't it be transit ABR instead of originating node, and also upstream LSP instead of original LSP.

2. Is there some special software entity besides the signaling entity (RSVP-TE or CR-LDP) at ABR that splices upstream LSP section to downstream LSP section. Is this entity what you call routing process in step 4 of section 4.4. Is this entity responsible for intercepting LABEL_REQUEST and PATH (as well as other CR-LDP and RSVP) messages at ABRs.

3. About crankback. It is stated that crankback is to the previous nearest ABR from the point of failure. At that ABR the routing algorithm by routing process is repeated.  Is there any mechanism that prevents the selection of the same ABR as was used previously. For example, refering to the following example

Area 1                      Area 0                    Area 2
--------------------    ------------------     ------------------------
|                  |    |                 |    |                       |
|  |N1  PPS1       -------    PPS3       --------      PPS5      |N2   |
|  |-------------- |ABR1  |--------------|ABR 2 |--------X-------|     |
|  |-------|       -------               -------- --|PPS7 |------|     |
|   PSS2   |       -------    PSS4       --------   |-----| PSS6       |
|          --------|ABR 3 |--------------|ABR 4  |---------            |
|                  -------               --------                      |
|                  |    |                 |    |                       |
--------------------    -------------------    -------------------------

Let's assume PPS5 failed and we initially cranked back to ABR 2 but there was no intra-area route to the destination so we cranked back to ABR1. At ABR1 when performing the algorithm of section 4.4 we should not consider ABR2 as this is from where we cranked back and we know that ABR2 already tried unsuccessfully to route to the destination.

Also, routing in response to crankback really depends on whether the failure causing the crankback was within the current area or another downstream area. If it was within the current area then same downstream ABR can be selected. If it was outside of the current area then a different ABR should be selected. For example, if PPS5 fails and we crank back to ABR2 then to ABR1 (as I eluded to above) then ABR2 should not be considered when routing at ABR1. On the other hand, if PPS3 fails and we crank back to ABR1 then ABR2 should be considered.

4. Have you given some thought to the idea that an LSP section failure in one area should not have an effect on surrounding LSP sections in other areas. For example, refering to the above picture, if PPS3 fails then another intra-area route through Area 0 between ABR1 and ABR2 is computed and LSP section reestablished along that route without affecting LSP sections PPS1 and PPS5. This would be similar to "local repair" where local = area.

Cheers,

Darek