Comment 16 for bug 1567655

Added the performance tag to this because even when neutron is up, when listing a large number of servers, we're getting ports for each server to get the security group information for each server, which is a lot of traffic to neutron. We should be smarter about this and cache the security group information per server like we do in the instance network info_cache for ports. We could heal the cache during the normal _heal_instance_info_cache periodic task.