some questions on virtual machine bridging.

[Andrew Gallatin]

On 05/28/12 12:12, Luigi Rizzo wrote:
> I am doing some experiments with implementing a software bridge
> between virtual machines, using netmap as the communication API.
>
> I have a first prototype up and running and it is quite fast (10 Mpps
> with 60-byte frames, 4 Mpps with 1500 byte frames, compared to the
> ~500-800Kpps @60 bytes that you get with the tap interface used by
> openvswitch or the native linux bridging).

That is awesome!

>    - and of course, using PCI passthrough you get more or less hw speed
>      (constrained by the OS), but need support from an external switch
>      or the NIC itself to do forwarding between different ports.
>    anything else ?

In terms of PCI passthrough / SR-IOV there are the emerging/competing
EVB and VEPA standards to allow VM<->VM communication to
go on the wire to a "real" switch, then back to the correct VM.

> * any high-performance virtual switching solution around ?
>    As mentioned, i have measured native linux bridging and in-kernel ovs
>    and the numbers are above (not surprising; the tap involves a syscall
>    on each packet if i am not mistaken, and internally you need a
>    data copy)

You should probably compare to ESXi.  I've seen ~1Mpps going to or from
from 1..N VMs and in or out a port on a 10GbE interface with ESX4
and newer.

Drew