RE: Moving LSP ownership between control plane and management plane

["Dino Bramanti" <Dino.Bramanti@marconi.com>]

Dimitri,

as Diego says in his post, the main use for the procedure defined in our ID
is for MP->CP handover.
This at least could be its immediate possible use and surely it is the
reason why we started thinking
about a similar mechanism within CP. But we have proposed the complete
procedure including the reverse CP->MP handover for several reasons.
The first one is related to the need of reverting back what MP-CP procedure
does in case of some kind of failure during its action. In addition, the
possibility of doing a CP->MP handover of LSP, leaving untouched
corresponding data plane state, could be useful for specific needs. An
example could be a Control Plane software upgrade to be performed on
network elements that are in service and carrying traffic over data plane.
In such a case it could be possible to perform a CP->MP handover of a
number of LSP, transfering their ownership to MP for the time required.

Furthermore, taking Stephen's suggestion about it and looking at the
proposed mechanism (i.e. handover flag usage within GRSVPTE signaling flow)
as a general way to act on CP state without affecting data plane state, the
need for a two-way procedure is required to provide it with maximum
flexibility.

Thank you
dino




>diego


>as part of the motivation section of this document it would be indeed
interesting to be much more explicit about the CP->MP >handover purpose(s)
before progressing on this part of the mechanism


>- d.