[one-users] cpu overcommit with kvm

[Giovanni Toraldo]

Il 04/05/2011 18:39, Ruben S. Montero ha scritto:
> You can guide the overcommitment by using the CPU attribute of the
> template. For example if you want to put 16 VMs in nebula02 with 8
> cores, just define the VMs with
> 
> CPU = 0.5

I already tried CPU setting, but the guest will see half of the real
frequency of my host CPU: this isn't cpu overcommit, I am just saying to
allocate half of my 8 cpu to that VM (or not?).

What I am trying to achieve is what I already do with other KVM-based
virtualization platforms:

allocating to every VM much more cpu than actually available on the host
(ex: 16 VMs with VCPU = 4 and CPU = 1), so a single VM can get a lot of
cpu power when needed.
In the case there are many VM using a lot of cpu, it will be equally
shared among them by the Linux process scheduler on the host.

Am I missing something? Is this a bad practice?

Thanks.

-- 
Giovanni Toraldo
http://www.libersoft.it/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 900 bytes
Desc: OpenPGP digital signature
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20110505/3955698a/attachment-0003.pgp>