Your Browser is not longer supported

Please use Google Chrome, Mozilla Firefox or Microsoft Edge to view the page correctly
Loading...

{{viewport.spaceProperty.prod}}

Reducing the multiprocessor level

&pagelevel(5)&pagelevel

Keep the number of the VMs' virtual CPUs (the multiprocessor level) as low as possible (see section "Number and multiprocessor level of the VMs"):

  • Specify the multiprocessor level of a VM. Do not simply use the multiprocessor level of the server.
  • When selecting the multiprocessor level of a VM, ensure that it is only large enough to permit the required (peak) load to be processed.
  • If possible, distribute the load over more than one VM with lower multiprocessor levels.
    In measurements on a /390 server with 4 CPUs a load was initially distributed over 2 guest systems each with 4 CPUs, and then over 4 guest systems each with 2 CPUs. It became apparent that the load distribution over 4 guest systems each with 2 CPUs caused around half the VM2000 overhead.
  • Do not start any guest systems "in reserve". Such guest systems cause VM2000 overhead even when they bear no load.
  • You can configure guest systems with an appropriate multiprocessor level to handle load peaks. However, detach the CPUs in the guest system in times when there is no load peak (/DETACH-DEVICE).