[VirtualNewbie]

Hi ,

We have an LPAR that is constantly accesing up to 600% of its entitlement capacity.

We have a debate where some of us newbies believe that entitlement should match the cpu that is constanlty being used vs. the experienced ones that says that the time that takes for the LPAR to aquire the extra capacity is too little to represent a performance problem.

I find it very hard to believe, hence I am asking all of you to either correct me or give me ammo to pursue what is right for my server.

Thanks in advance for the help...

[rydekull]

I'd measure its average CPU utilization and let it have that value as its required value.

Then I'd weight it depending on how critical it is compared to the other LPARs on said system.

[ross.mather]

Depends on whether you are using CPU pooling or not....

If every LPAR has access to a pool of CPU resources then iover capacity is not a problem so long as the overall usage of the Pool of resources is not presenting a problem. Acquiring of the cycles is not a time consuming issue. The issue is only id the 600% CPU capacity always available when you need it, and if not what happens when it isn't?

If on the other had the principle is that each LPAR has its own fixed allocation of CPU and extra cycles are given on a "goodwill" basis then you want to demand additional resources.

As a personal rule of thumb, 600% of 0.1 CPU is 0.6 CPU, but 600% of 4 CPUS is 24 CPUs and thats a bit of a difference, but I never want to be reliant on more than 1 CPU being borrowed from another location.