[jrigler]

I am trying to integrate lparstat into my system optimization solution for our 5.3 lpars. The stats that seem to add new information to what I already know from vmstat:

vcsw - virtual context switches
phint - phantom interrupts

The man page leaves something to be desired on these options, I am guessing that one or both of these represent cycles stolen from this system to be used by another system. However why isn't this 0 on a 510 that has only one lpar.

I want to be able to definitively show my users that if I take X number of cycles from their lpars, that 'sharing' actually works and unless they all click on 'run' at the exact same time, cycles between the lpars are shared and they can all peacefully coexist with a smaller processing footprint.

[Jorke]

Use topas and press L

This shows the sharing off physical cPU's between LPARS

Jorke
Certified system expert