The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Mar> msg00365



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Comments on draft-ip-optical-framework-00.txt

  • From: Tony Li <tony1@home.net>
  • Date: Tue, 28 Mar 2000 00:48:31 -0800
  • CC: Krishna Bala <kbala@tellium.com>, Jonathan Lang <jplang@lux.chromisys.com>, curtis@avici.com, Jagan Shantigram <jagan@photonex.com>, Khaled Elsayed <khaled@ieee.org>, mpls@UU.NET, ip-optical@lists.research.bell-labs.com
  • Organization: Li Consulting

> You may be right that both models can support all services provided
> signaling and routing are IP-based.  However, there are two fundamental
> differences between the Open and the Peer model: (1) In OPen model if an
> operator wants to offer "optical Pipe" services regardless what svc is
> carried within the pipes the operator would not like to share all its
> network resource information to its clients.  Therefore what is needed for
> information exchange between the client NE and the optical network is a
> standard interface defining the minimum set of required parameters without
> sharing optical netrwork resource info with the clinet NE.  That is what UNI
> spec being developed by OIF. (2)We do not need to impose the routing and
> signaling protocols to every client network elements for providing the
> optical layer services.  Al they need to adhere to is the simplie UNI spec.
> Rest is done by the optical layer using IP signaling and routing.  The
> driver for using IP routing and signaling in the optical layer is simply
> because these are well developed and well suited for the optical layer
> functions and there is no need to reinvent the wheel.  Also, when
> multi-vendor optical networks are to work together other protocols such as
> BGP can be easily adopted.

Based on this, the difference is that you're emphasizing is the use of another
protocol at the edge of a service providers network.  If this is an accurate
assesment of your position, then it would behoove us to distinguish between the
intra-domain model and the inter-domain model.

For example, one could reasonably say that the MPLS/TE model is currently a peer
intra-domain model and that we've completely blown off the inter-domain model.
So far, this seems to have a reasonable amount of appeal to the customer base.


> S0 unless we assume that all services to be provided by an optical network
> (inclcuing pipe svc) are accessed via routers that are within the opretor's
> domain the peer model by itself is unworkable and once we have the open
> model the need for peer model is diminshed to almost naught.

If I follow that logic, it is impossible to get any services if multiple domains
exist.  I think the Internet is a small counterexample.

Regards,
Tony