Re: [mpls] MPLS ring protection reconsidered ??

[Huub van Helvoort <[e-mail]>]

Hello Loa,

Thank you for your analysis.

The key source of information required for protection switching
in a ring topology are the "ring-maps" present in each node in
the ring.
A "ring-map" contains the identity and location of each "ring-node"
in the ring. It also contains information of the "ring-LSPs".

draft-kompella-mpls-rmr describes how to fill the "ring-maps"
and can be extended to multiple rings sharing a link.
draft-cheng-mpls-tp-shared-ring-protection describes the messaging
mechanism to provide the protection switch in single and multiple
rings.

I support your proposal below.

Regards, Huub.
=============

 > As far back s the 73rd IETF in Minneapolis John and Adrian made
> a report on "Requirements for Ring Protectionin MPLS-TP". The
> conclusions were that we could do topology specific protection
> solutions if the benefits are big enough.
>
> Such solutions need to meet the same requirements as linear
> protection and it has to be show that it can't be done by linear
> protection only.
>
> At that time we did not see that there were things that would not
> be as readily done by the linear protection being specified at that
> time.
>
> Today we have to drafts that address ring topologies, one
> draft-kompella-mpls-rmr addresses Resilient MPLS Rings in an MPLS-TE
> environment. The other draft-cheng-mpls-tp-shared-ring-protection
> addresses protection in an MPLS-TP environment.
>
> Both recognizes that ring topologies are very common and that very
> efficient mechanism for keeping traffic flowing in case of failures
> are possible to design. Sometime far better than what is the case if
> the actual ring topologies are view as a linear topology,
>
> The first document (draft-kompella- ) looks primarily on the operations
> within a single ring and how fast and simple mechanisms for protection
> can be deployed. A ring topology is a very common deployment scenario.
> While, the draft-kompella from a solutions point is somewhat orthogonal
> to draft-cheng, it does also discuss the dynamic control plane for mpls
> ring, including auto-discovery and signaling. It seems that there are
> opportunities for co-operation between the two drafts in this area.
>
> The other (draft-cheng- ) looks at what is called MPLS shared ring, i.e.
> a rather high number can shared the same path around the ring, and all
> traffic can be protected by a single operation.
> Another aspect of the shared tunnel is that if part of the ring
> (typically 2 nodes and one link) are part of more than one ring. It
> becomes possible to protect against more than one failure.
>
> Maybe it is time to revisit the question and see if we want to adopt
> working group documents for the two scenarios outlined above.
>
> /Loa


-- 
*****************************************************************
               请记住，你是独一无二的，就像其他每一个人一样