[openstack-dev] [quantum] Blueprint Announcement: OVSPlugin Support for Hardware Devices
satish at aristanetworks.com
Thu Jan 31 01:20:22 UTC 2013
> > Hi Gary
> > The difficulty to do it at a layer above the plugins is that the network
> > segmentation decisions (VLAN id allocator or Tunnel ID allocation are done
> > by individual plugins today). We view the driver as an extension of the
> > plugin and extends the capabilities of the plugin and allows it to manage
> > more complex deployments like TOR switches in the example.
> > I feel we should certainly discuss this in the light of Bob's modular L2
> > (ml2) plugin architecture and see how best to layer it.
> Satish, I agree that there must be some coupling between plugins and driver
> interfaces (e.g., it only makes sense to use a VLAN driver interface with a
> plugin that uses VLANs for segmentation). But perhaps what Gary is saying
> is that it would make sense for the VLAN driver API and driver
> implementations to exist elsewhere in the code tree, so it would be cleaner
> if other plugins decided to adopt and use that same VLAN driver API (again,
> just as an example).
Yes I guess that makes sense. But just to be sure I understood right - this level of separation or abstraction is possible only with something like the proposed ml2 framework (as an example) - not the current plugin framework (or am I missing something) ?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the OpenStack-dev