Re: [CCAMP] Question on LSP diversity

[Gert Grammel <[e-mail]>]

Hi Matt,

I recall asking a different question, namely what kind of separation is represented by the domain boundary and where it is described. I understood that the definition of the boundary is not in the scope of ccamp but the draft(s) should describe which assumptions are made about the nature of the boundary.

About the cases themselves (using the terms "client" "server" in a control plane sense as described in Daniele's presentation) I try to point out some networking scenarios with the aim to spell out the scenarios:


-          1. Single homing --> it should be spelled out what is understood to be single homing

§  Would a client connected with two fibers into the server domain qualify for single homing?

§  Would a client connected with two fibers into the server domain, each attached to different server domain routers still qualify for single homing?

§  How do we account for a client LER which is connected to two client LSRs whereby one connection stays inside the client network and one connection is connected to the server network? The client LER is dual homed  from a client perspective but from a server perspective it appears single homed.

-          2. Dual homing --> it should be spelled out what is understood to be dual homing

§  Would a client connected with two fibers into the server domain, each attached to different server domain routers qualify for dual homing?

§  As we are talking about routers it is safe to assume that some are multi-homed where homing is >2. Unless there is a reason to limit the case to exactly dual homing we should name this 'Multi-homing' instead

§  Is a client which is connected to two distinct server domains considered dual homed or multi-domain?

-          3. LSPs with different edge and core nodes --> the content of this case probably depends a lot how single/multi homing is defined

§  Does this cover cases where one end is dual homed and the other end single homed? Or is this covered above?

§  Does this cover also cases where n clients are connected to m server nodes with n>m?

-          4. Multi domain

§  Does this mean multiple interconnected core domains, or does it refer to a client connected to several server domains?

§  Is it related to a stacked client-server-server case?

During his presentation George pointed out that he is considering "All networks under a single administration" when presenting http://tools.ietf.org/agenda/89/slides/slides-89-ccamp-4.pdf . This seems to suggest that there is also demand to work on single domain cases. Not sure how much merit is in that, but perhaps worth as a reference

Best

Gert



From: CCAMP [mailto:[e-mail]] On Behalf Of Matt Hartley (mhartley)
Sent: 06 March 2014 14:50
To: Daniele Ceccarelli; CCAMP ([e-mail])
Subject: Re: [CCAMP] Question on LSP diversity

Daniele,

ISTR Gert asked about this on Tuesday, and the reply was that this is outside the scope of CCAMP and therefore of the work being proposed here.

I'm sure someone will correct me if I'm wrong about that :)

Cheers

Matt

From: CCAMP [mailto:[e-mail]] On Behalf Of Daniele Ceccarelli
Sent: Thursday, March 06, 2014 6:36 AM
To: CCAMP ([e-mail]<mailto:[e-mail]>)
Subject: [CCAMP] Question on LSP diversity

Hi,

bringing this to the list due to time constraints.

We said we have 4 use cases for LSP diversity:

-          1. Single homing

-          2. Dual homing

-          3. LSPs with different edge and core nodes

-          4. Multi domain

In cases 3 and 4 there is the need for "coordination" between edge nodes. Is such coordination in CCAMP scope or we assume that "somehow" the two edge nodes exchange such piece of info (e.g. NMS, the triangle with the eye, or whatever)?

BR
Daniele