Configure virtual CPU capping

To limit the amount of CPU resources a virtual machine (VM) can use, you can configure virtual CPU (vCPU) capping. vCPU capping can improve the overall performance by preventing excessive use of host’s CPU resources by a single VM and by making it easier for the hypervisor to manage CPU scheduling.

Procedure

  1. View the current vCPU scheduling configuration on the host. 
    # virsh schedinfo <vm_name>
     
    Scheduler      : posix
cpu_shares     : 0
vcpu_period   : 0
vcpu_quota    : 0
emulator_period: 0
emulator_quota : 0
global_period  : 0
global_quota   : 0
iothread_period: 0
iothread_quota : 0
  2. To configure an absolute vCPU cap for a VM, set the vcpu_period and vcpu_quota parameters. Both parameters use a numerical value that represents a time duration in microseconds.
    1. Set the vcpu_period parameter by using the virsh schedinfo command. For example: 
      # virsh schedinfo <vm_name> --set vcpu_period=100000

      In this example, the vcpu_period is set to 100,000 microseconds, which means the scheduler enforces vCPU capping during this time interval.

      You can also use the --live --config options to configure a running VM without restarting it.

    2. Set the vcpu_quota parameter by using the virsh schedinfo command. For example: 
      # virsh schedinfo <vm_name> --set vcpu_quota=50000

      In this example, the vcpu_quota is set to 50,000 microseconds, which specifies the maximum amount of CPU time that the VM can use during the vcpu_period time interval. In this case, vcpu_quota is set as the half of vcpu_period, so the VM can use up to 50% of the CPU time during that interval.

      You can also use the --live --config options to configure a running VM without restarting it.

Results

  • Check that the vCPU scheduling parameters have the correct values. 
    # virsh schedinfo <vm_name>
     
    Scheduler      : posix
cpu_shares     : 2048
vcpu_period    : 100000
vcpu_quota     : 50000
...