Hi,
Just to clarify.
Let`s say that a VM has 3 vcpus and start on core 0 and core 1.During the VM life the vcpus will use only core 0 and core 1 or also can use other available cores from server?
Is the hypervisor that will give a certain core to a certain vcpu from VM to run?
Every time when a process need to run on a certain vcpu then the hypervisor will give a certain core where this vcpu run.right?
tnx
No.
First of.. if the GuestOS place workload on every of their 3 vCPUs the hypervisor needs to find 3 free logical CPUs (pCore and HT counts). If no 3 free pCPUs are find the complete VM have to wait. With a modern GuestOS the hypervisor report to the VM that it is a virtual and the GuestOS may try to place process on less CPUs instead of spreading them around. The hypervisor can detect vCPU without workload and then it doesnt need to find a pCPU for.
As long as you not have pinned a vCPU to a given pCPU they moved around in your multi Core CPU Socket/System. The Hypervisor try to place the vCPU within the NUMA boundaries of your System if possible or depending on the VM configuration.
Regards,
Joerg
Just for sure. This book can give you good answeres Rubrik | eBook: VMware vSphere 6.5 Host Resources Deep Dive
hi,
tnx.i like short answers to my questions.just for sure there are no answers in this book.
No.
First of.. if the GuestOS place workload on every of their 3 vCPUs the hypervisor needs to find 3 free logical CPUs (pCore and HT counts). If no 3 free pCPUs are find the complete VM have to wait. With a modern GuestOS the hypervisor report to the VM that it is a virtual and the GuestOS may try to place process on less CPUs instead of spreading them around. The hypervisor can detect vCPU without workload and then it doesnt need to find a pCPU for.
As long as you not have pinned a vCPU to a given pCPU they moved around in your multi Core CPU Socket/System. The Hypervisor try to place the vCPU within the NUMA boundaries of your System if possible or depending on the VM configuration.
Regards,
Joerg