HP-UX vPars and Integrity VM V6.3 Administrator Guide

5 CPU and Memory
5.1 Configuring CPU resources for VM guests
5.1.1 Processor virtualization
VM guests are configured with virtual processors. A vCPU is a virtualized schedulable entity. Virtual
processors are mapped to physical CPU, cores as a part of VM guest scheduling. For the purpose
of this discussion, the term “physical CPU” refers to a processing entity on which a software thread
can be scheduled. Each vCPU is independently scheduled as a single thread of execution on a
physical CPU, subject to the entitlements discussed in Section 5.1.2 (page 53). The scheduling of
all vCPUs belonging to a guest and across guests is independent of each other. This helps in
maximizing the utilization of physical CPU resource across many vCPUs belonging to different VM
guests.
Each VM guest has at least one vCPU. Use the hpvmcreate -c number_vcpus command to
specify the number of virtual CPUs that the VM guest can use. The maximum vCPU count that can
be set for a VM guest is 32. If you do not specify the number of vCPUs, the default is 1.
For example, to set the new VM guest vmguest1 to have two vCPUs, enter the following command:
# hpvmcreate -P vmguest1 -c 2
A running VM guest cannot use more vCPUs than the number of physical CPU cores on the VSP
system. Do not set the number of vCPUs higher than the physical number of CPU cores, as this can
prevent the VM guest from starting.
You can change the number of enabled CPUs in VM guests running HP-UX, using the hpvmmgmt
–c num command from within the guest OS. This command sets the number of enabled virtual
CPUs to the number indicated by num (up to the number of CPUs the guest is booted with), and
disables the others. Disabled virtual CPUs do not show up in the guest when you run commands
such as top or GlancePlus, and do not consume resources on the VSP. However, disabled
virtual CPUs still appear on the VSP, for example when you run the hpvmsar command.
NOTE: HP Integrity VM does not support running real-time applications on the guest. Scheduling
and precise timing properties that can be relied upon on physical hardware are not guaranteed
to be preserved in a VM guest. In particular, changing the hires_timeout_enable(5) HP-UX
tunable might not have the desired effect.
5.1.2 vCPU entitlements
Entitlement is the minimum amount of processing power guaranteed to the VM guest from each
virtual CPU. When you create a VM guest, you can use the hpvmcreate -e command to specify
the entitlement as a percentage, a value between 5% and 100%. If you do not specify the
entitlement, the VM guest receives 10% entitlement by default.
When the VM guest starts, the VSP ensures that sufficient processing power is available for every
running VM to receive its entitlement. For VM guest with multiple virtual CPUs, the entitlement is
guaranteed on each vCPU in the VM’s configuration. For example, if a VM guest has four vCPUs,
and the entitlement is set at 12%, the VSP ensures that the equivalent of at least 48% of one physical
CPU is available to that guest.
To allow multiple VM guests to run at the same time, make sure that the entitlement of each VM
guest does not prevent the other VMs from obtaining sufficient processor resources. The sum of all
entitlements across all active VM guests cannot total more than 100% for any physical processor.
If a VM guest is busy, and sufficient physical CPU resources are available on the VSP system, it
can receive more than its entitlement. When there is contention for processing resources, each VM
is limited to its entitlement.
5.1 Configuring CPU resources for VM guests 53