Re: Network modeling & NML group in the OGF

[Freek Dijkstra <fdijkstr@science.uva.nl>]

I got some off-list replies. Let me elaborate a bit.

The network markup language (NML) working group in the OGF -in my view- 
seems focussed on network descriptions for multi-domain scenarios. In 
those scenarios loose coupling and extensions to refer to other 
resources are important. GMPLS seems mostly focussed on a control plane 
within a single administrative domain.

Nevertheless, I see overlap. I imagine a future where GMPLS is used 
within a single administrative domain, that data is extracted and in 
abstracted form announced to select partners for multi-domain lightpath 
provisioning.

Such a tool would be a lot easier if there was a common network model. 
Translating between two syntaxes is easy. Translating between two models 
is very hard.

In an ideal world, every technology uses the same model, so conversion 
is easy. We don't live in an ideal world. For example, GMPLS does NOT 
follow G.805 and G.8080. To give two examples: Fundamental concepts in 
GMPLS are labels and capacity (bandwidth). Neither of these concept is 
is present in G.805. In GMPLS, there is no 1:1 relation between 
encapsulation "layers" and switching capability "layers". In GMPLS, 
there is (link connections and subnetwork connections are on the same 
layer network).

Because of this difference in model, it is hard to convert from a set of 
TLV in OSPF-TE and a network description in G.805 functional elements or 
vice versa.

What I would hope for the NML working group is to learn as much as 
possible from all these different models, and decide on one which is 
most usable for all, and is easy to convert to other models and syntaxes.

Clearly, G.805 and G.8080 are input, but they are not complete (*). The 
model (if any) used in GMPLS is another input, and I am seeking that 
input from this list.

Regards,
Freek


(*) For example, we like to add time-based information to track changes 
in the network topology. We like the label and switching capability 
concepts in GMPLs, neither of which is in G.805 or G.8080. Also you 
don't want to know how many people assume that a device interface maps 
to a single G.805 "connection point" -- that is not true, since a device 
interface is a physical interface, which may map to multiple logical 
interfaces, one for each layer. These things need to be specified.


PS: It was perhaps not completely clear in my previous mail. We (the 
developers of the network description language, NDL) use RDF. However 
that this is not always true for everyone in the Network markup language 
(NML) working group in the OGF. Most of the other partners prefer to use 
XML for example. We decided to simply come up with a common model in 
UML, and allow multiple syntaxes. Again, if there is a common model, 
translating between syntaxes is trivial. That common model is the hard part.