[ironic] : Third-party ML2 plugins for Openstack ironic with virtual networks

Is anyone using virtual networks for their Openstack Ironic installations? Our flat network is now past 3000 nodes and I am investigating Arista's ML2 plugin and / or Mellanox's NEO as the ML2 plugin. In addition to scaling we also have additional requirements like provisioning a bare-metal serverin a conference room away from the DC for demo purposes. I have general questions on whether anyone is actually using the above two (or any other ML2 plugins)with their Openstack ironic installations ? Thanks,Fred.