Hello,

I know that "For instances with a floating IPv4 address, routing between self-service and provider networks resides completely on the compute nodes"[1] when enable neutron dvr. This can eliminate performance issues with network nodes.

But it need fix floating ip to vm. I want to ask when employing floating IP port forwarding, is it feasible to enable direct access to the compute nodes for traffic?

I may encounter a scenario where tens of thousands of ports are forwarded from a floating IP to different ports on different VMs. This would result in a significant concentration of traffic at the network node. I would like to inquire if there are any methods available to alleviate the pressure on the network node or to directly distribute the traffic across the compute nodes.

Appreciate for any assistance or suggestions.

Best regards, Han Guangyu

[1] https://docs.openstack.org/neutron/train/admin/deploy-ovs-ha-dvr.html#deploy...