[ovs-discuss] Switch distribution over hardware

Ben Pfaff blp at nicira.com
Sat Jan 23 17:55:08 UTC 2010

On Sat, Jan 23, 2010 at 08:32:34AM +0000, DarkBls wrote:
> so far, as I know here are the way to distribute OvS among several
> hardware (with supported OvS feature):
> centralized controller with OpenFLow
> Ethernet Over GRE
> What about a solution ? la KVM/QEMU with multicast group ? It's a
> pretty nice way to takle the issue too. A virtual switch stacking with
> multicast.

It's a reasonable idea.  I don't think we at Nicira have any plans to
implement this right now, but I expect that we would accept well-written
patches for it.

For anyone here who doesn't know what DarkBls is talking about, here's
the appropriate section from the QEMU manpage:

       -net socket[,vlan=n][,name=name][,fd=h][,mcast=maddr:port]
           Create a VLAN n shared with another QEMU virtual machines using a
           UDP multicast socket, effectively making a bus for every QEMU with
           same multicast address maddr and port.  NOTES:

           1.  Several QEMU can be running on different hosts and share same
               bus (assuming correct multicast setup for these hosts).

           2.  mcast support is compatible with User Mode Linux (argument
               ethN=mcast), see <http://user-mode-linux.sf.net>.

           3.  Use fd=h to specify an already opened UDP multicast socket.


                   # launch one QEMU instance
                   qemu linux.img -net nic,macaddr=52:54:00:12:34:56 \
                   -net socket,mcast=
                   # launch another QEMU instance on same "bus"
                   qemu linux.img -net nic,macaddr=52:54:00:12:34:57 \
                   -net socket,mcast=
                   # launch yet another QEMU instance on same "bus"
                   qemu linux.img -net nic,macaddr=52:54:00:12:34:58 \
                   -net socket,mcast=

           Example (User Mode Linux compat.):

                   # launch QEMU instance (note mcast address selected
                   # is UML's default)
                   qemu linux.img -net nic,macaddr=52:54:00:12:34:56 \
                   -net socket,mcast=
                   # launch UML
                   /path/to/linux ubd0=/path/to/root_fs eth0=mcast

> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org

More information about the discuss mailing list