Re: draft-gan-fast-reroute-00.txt

[Ping Pan <pingpan@juniper.net>]

"Metz, Eduard" wrote:
> 
> haven't read the draft in detail yet, but adding support for SRLG to avoid
> detours to be routed over links that share the same risk seems very
> useful.Sometimes even a plain requirement to deploy fast re-route in a
> usefull way. As an example consider the situation described in section 2,
> when link 1-2 and link 1-3 run in the same cable both the main LSP and the
> potential detour are affected by a single cable-cut. Leaving you with
> regular re-route.
> 

Thanks for the input. Will get back to you asap.

> Some other questions:
> - What will happen if the detour from Node1 to Node3 (picture section 2)
> passes Node4? Will the detour terminate at Node4, or will the traffic will
> pass link 3-4 two times?

I think that it is OK to have detour paths traversing through the same
nodes that the main LSP goes through. This type of loops does not
present problems in packet forwarding... except perhaps inefficient use
of network resources. But we need to be careful with the loop detection
mechanism described in the RSVP-TE draft.

> - Similar to the above, but now for the case where the detour passes the
> Egress?
> 

No sure this can happen after CSPF computation.

Thanks!

- Ping