[collectd] : what does "virt_cpu_total" from libvirt mean exactly?

[Zhou, Yuantai]

Hi,
We are collecting the "virt_cpu_total" with libvirt from kvm and want to setup monitoring based on the virt_cpu_total for the VMs. 

We noticed that the value of the "virt_cpu_total" is over 100 for some of our VMs. If "virt_cpu_total" is a percentage, how does libvirt determine the virt_cpu_total? Does it take number of sockets\cores\threads into consideration or simply run a "top" command to get the CPU usage? 

We are using collectd version 4.10 with default cpu model. Wondering if newer version has different implementation so the "virt_cpu_total" is based on total available CPU resources with a value between 0-100%? Or if we set additional feature flags, and\or the CPU topology (sockets/cores/threads), can we get relative virt_cpu_total against total cpu resources so it would be 0-100%?

Thanks.
Yuantai