docs/doc/source/node_management/openstack/pci-passthrough-ethernet-interface-devices.rst

PCI Passthrough Ethernet Interface Devices

For all purposes, a passthrough interface behaves as if it were physically attached to the virtual machine.

Therefore, any potential throughput limitations coming from the virtualized environment, such as the ones introduced by internal copying of data buffers, are eliminated. However, by bypassing the virtualized environment, the use of passthrough Ethernet devices introduces several restrictions that must be taken into consideration. They include:

• no support for , , , or host interface monitoring
• no support for live migration

partner

start-after

avs-bullet-3-begin

end-before

avs-bullet-3-end

starlingx

A passthrough interface is attached directly to the provider network's access switch. Therefore, proper routing of traffic to connect the passthrough interface to a particular project network depends entirely on the tagging options configured on both the passthrough interface and the access port on the switch.

partner

start-after

avs-text-begin

end-before

avs-text-end

The access switch routes incoming traffic based on a ID, which ultimately determines the project network to which the traffic belongs. The ID is either explicit, as found in incoming tagged packets, or implicit, as defined by the access port's default ID when the incoming packets are untagged. In both cases the access switch must be configured to process the proper ID, which therefore has to be known in advance.

Caution

On cold migration, a passthrough interface receives a new address, and therefore a new eth x interface. The IP address is retained.

In the following example a new virtual machine is launched by user user1 on project project1, with a passthrough interface connected to the project network net0 identified with ID 10. See Configure PCI Passthrough ethernet Interfaces <configure-pci-passthrough-ethernet-interfaces>