[libvirt-users] performance tradeoffs/penalty

Felicitus felicitus at felicitus.org
Wed Apr 10 12:45:02 UTC 2013


> does libvirt make a user pay penalty, even if only a small amount, for
> using it when compared to raw qemu-kvm command line?

Not that I know of, except that the libvirt daemon uses a few additional
CPU cycles. Technically yes, in the spirit of the question no. Have a
look at your process tree; you'll see command lines in the same way as
you'd expect them to be when you start machines manually.

> if it does where the overhead goes? quest or host, or maybe both?

No additional CPU cycles for the guest, a few additional CPU cycles for
the host to get status information and expose it via the daemon.

> if overall performance is slower then how small it is comparing to all
> the advantages it offers?

It is completely negligible.

Felicitus




More information about the libvirt-users mailing list