I would like to be able to give, for example, 33% for dom1 and dom2 and 33% for dom0, but I know that we can only work with VCPUs.
It doesn't work like this. Guests don't normally consume CPU continuously (well, unless you're running something like numerical simulations, which I really wouldn't recommend for virtualization).
First of all, if you do nothing at all then guests' virtual CPUs are scheduled to run on physical CPUs at random. When a guest starts to do something, a physical CPU is picked and the guest runs on that for some short timeslice. You can easily have more virtual CPUs than physical CPUs. If your guests are mostly idle, the this is fine.
Now you can also pin a virtual CPU to a subset of physical CPUs (or to just a single physical CPU). Using virsh/libvirt:
virsh vcpupin guest 0 1would pin guest's virtual CPU #0 to physical CPU #1. So now when guest's virtual CPU #0 wakes up to do some work, it will only be scheduled to run on physical CPU #1, if necessary waiting until physical CPU #1 becomes free.
If you had a quad-core machine (4 physical CPUs) and 1 dom0 + 3 guests, and each dom0/guest has a single virtual CPU, then you might consider pinning the dom0 and each guest to a separate physical CPU. Whether this is a good idea or not depends on what each guest will be doing and requires careful analysis and performance monitoring.
Use the command: virt-top -1 to show which guests are running on which physical CPUs in real time. Rich. -- Emerging Technologies, Red Hat - http://et.redhat.com/~rjones/ Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE, United Kingdom. Registered in England and Wales under Company Registration No. 03798903
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature
-- Fedora-xen mailing list Fedora-xen@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-xen
