[Nml-wg] Network laydowns

[Freek Dijkstra]

Connah Jessica J wrote:

> I am currently working on a project looking at capturing networks in a
> tool agnostic way so that we can then easily migrate the networks onto
> different tools.  I am interested in the outputs of your working group
> and what you decided was the best way to tackle this problem.

Hello Jessica,

The working group has not reached the stage of integration with existing
tools (in fact, it is only chartered to come up with a schema and syntax
for network topology representation). A few software implementations are
in the making, but as far as I know, those are all stand-alone.

That said, most of the group members have encountered the same problem
you are facing. Perhaps they can point you in the right direction. In
particular, one of the earlier contributors developed the cNIS tool,
which was designed with this interoperability problem in mind. cNIS was
developed in the EU-funded GÉANT2 project.

I can relate a bit about my experiences in this field. Originally coming
from academy, our solution was to come up with a tool-agnostic network
description, and build custom tools that import/export to that syntax.
The network description we developed was NDL, and it was based on the
ITU-T G.800 standard. G.800 provides building blocks to model networks
in a technology independent way. The outcome of the NML working group
will also be based on G.800, although some changes have been made, using
the experience from other group members (NML is basically
unidirectional, and there is a generic way to describe relations between
the building blocks).

One of the lessons we learned is that converting between syntax is easy,
as long as the underlying model is the same. Using the G.800 building
blocks and the NML schema (which, when finished, describes how to use
those building blocks for particular technologies), there is a common
model. Translating between syntax (e.g. in our working group there is
voice to create a XML and a RDF syntax) is then relative simple. Four
important choices for the model are (1) what identifiers to use for each
component, (2) if the basis is unidirectional or bidirectional, (3) if
the model is a single layer or a multi layer (cross-technology)
description, (4) if the described network is single domain or
multi-domain (thus often distributed and aggregated).
The choices our group made where: (1) we use URI for all identifiers;
(2) the basis is unidirectional (where bidirectional and broadcast are
described using unidirectional -- a known limitation is that this does
not work well for cellular networks); (3) we explicitly focus on
multi-layer network, and how to relate the adaptation between the
different technologies; (4) the working groups focus on the exchange of
topology information between domain, and hence is multi-domain.

Hope this helps a bit. Perhaps you could explain a bit about your
particular problem. That may give a clue to other group members in your
problem.

Regards,
Freek Dijkstra