hvm / qemu performance tuning

[cirr%sdf.org@localhost]

I've recently had to move from PV DOMUs to HVM DOMUs due to a particular
OSes drop of support of native Xen.  While under PV performance was great,
but under HVM the is a noticeable performance hit especially with each
DOMU requiring qemu.

According to:  https://wiki.netbsd.org/ports/xen/howto/

stub domains
Some (Linux) dom0 systems use something called "stub domains" to isolate
qemu from the dom0 system, as a security and reliabilty mechanism when
running HVM domUs. Somehow, NetBSD's GENERIC kernel ends up using PIO for
disks rather than DMA. Of course, all of this is emulated, but emulated
PIO is unusably slow. This problem is not currently understood.

Are there any Xen parameters for the DOMU that can be used to reduce qemu
over head? For instance I'm using:

vnc = "0"
nographic = "0"
serial = "pty"

-videoram gets set to 4, can that be eliminated as well?

Based on what is stated in the wiki address above, is there any work being
done in this area or would it be best to run the DOM0 under Linux for Xen
servers that require many HVM guests?


Thanks!