Hi,
On 4 Mar 2012, at 13:55, Freek Dijkstra wrote:
Hi all,
Last Friday Jerry and I had a short conversation. Jerry's idea is to
define recursive topologies, as he described in his slides to the list.
He convinced me that this is a great concept, as it allows for simple
topology abstraction.
While I wholeheartedly support the idea, I like to describe the concept
before giving a concrete proposal for changing the NML schema.
1. Connecting the (sub)topologies
2. Hierarchical identifiers
3. Updates and Versioning
Feedback is highly appreciated.
1. Connecting the (sub)topologies
I agree with this idea, and the option of defining aliases. This makes for a more stable inter-domain topology.
I would suggest that we allow abstracted topologies to contain some hints regarding internal availability.
I don't mind doing something like this as long as it is an optional
announcement by the local network. Thus a network can announce what
they are comfortable announcing. And an external path finder agent
can gather as much topology as they can acquire. I think some of
the internal availability announcements become less important if we
modified the Reservation Request to more acurately reflect the
desired constraints - in particular to loosen the endpoint
speciications to allow some selection criteria on a set of available
endpoints. But this latter approach is an issue for NSI, not
really NML.
3. Updates and Versioning
In the Automated GOLE experiments I've found that being able to see a version of the topology used would be a nice feature. Currently the demo still uses a centralized topology, so knowing the used version is more important. Still, knowing when your neighbor last updated their topology can help debugging issues. So I think it would be a nice feature to have, although not necessarily something that we'd include in the schema itself though.
I concur that we need to transition the topology management to the
networks/GOLEs themselves as soon as we can - this is the right
thing to do. But to do that, we should also have a well considered
plan for how we expect NSAs to construct their world views - and
have all NSA developers implement this as part of the transition.
There is no urgent rush - better to think it though first. I
would vote for the basic approach that each NSA builds a world view
from direct neighbors and then they announce their world view,
rather than each domain simply announcing only their own local
topology. I think the world view approach will scale better.
Thoughts?