Originally posted by: The_Doctor
I'm sure there are many interpretations of Minimum Capacity, but in simplistic terms, my interpretation is --> Minimum Capacity is the MINIMUM amount of CPU allotment that must be available in the PHYSICAL FRAME, otherwise the LPAR will not start.
Personally, I think you are confusing Minimum Capacity with Entitled Capacity. They are significantly different. Entitled Capacity (EC) is the CPU allocation that you are guaranteed to get during each time slice. So, with EC = 1.0 you are guaranteed to get 1.0 CPUs worth of processing during each time slice
BUT ONLY IF YOU NEED IT. Excess/unused cycles are generally ceded back to the hypervisor for potential distribution to other needy LPARs.
PC shows you what is
physically consumed during the interval. So, PC = 0.01 indicates an average of 1% of 1 CPU was used during that interval. A PC = 3.45 would indicate an average of 3 CPUs at 100% + 1 CPU at 45% for a total average consumption of 345% during the interval.
The hypervisor recalculates changes every 10ms or 100 times per second (aka a time slice). What you're guaranteed in terms of EC does not change. Once an LPAR is powered on and the hypervisor sets the EC,
the EC stays at that value until you change it (either forcibly thru the use of DLPAR or similar changes thru the profile, which would require a POWER OFF/ON cycle). I therefore do not understand where any latency would come into play..... but maybe I'm missing your point.
You also wrote >
does xmtopas process make some role in such processing unit increase/decrease operations? I think it's safe to speculate & say the answer is -->
NO !#AIX-Forum