Open Stack

Kevin,

This is just a final top-post to say that I'm going to spend the next
week noodling over this discussion in more detail and trying to flesh
it out into a proposal.  I think we're pretty much in agreement about
where the problems are in the model.  It feels like we'll just be
beating a dead horse to continue to hash that out here.

I will take the action to write up a proposal based on this discussion
and have it posted before the 18th when we are planning to have a
discussion at the operators meetup.  I will focus on the L3 network
subnet group thingy, floating IPs, and how it relates to a group of L3
connected L2 segments.  I'll try to work out a migration plan both for
the DB and for existing work-loads.  I'll go back and review Assaf's
and others' comments on grouping the networks.  I will also consider
Neil's proposal for routed networks and how it fits in.  Maybe it fits
in better.

I will not focus much on the Nova scheduling aspect.  I acknowledge
that is an important aspect to solving this whole problem.  But, in my
mind, it hasn't changed much since we discussed it at the mid-cycle
and I think you had some interest in working it out.  Essentially,
there are still multiple segments modeled as neutron networks that
will be collected in to a group "somehow" and we have a way to pass
hints around like we discussed to schedule an instance and create a
port on the appropriate network.

Carl