>I would expect PROJECTCPU to be pretty accurate as I would expect it to 
>work off when the task is marked zIIP-eligible.

For non-z/OS core code, there is no such thing as a task marked 
zIIP-eligible (ignoring the zAAP on zIIP functionality).

So for products that dual-path "if zIIP available then use them, otherwise 
use tasks", projectCPU will not account for them.
For the purposes of projection accuracy, it would be better if the 
products used "if zIIP available or projectCPU is on then...".
That bit isn't currently a programming interface (SVT_CpuProjection) but 
could be made so.

>For example, time to do 
>"Expensive Work X" reduced by 64% by splitting (using educated 
>"guesstimation") the work up across three dispatchable units, including 
>overhead of startup/takedown of the two additional SRBs 

That would apply whether zIIPs were in use or not (although on a machine 
where offload engines are faster than standard CPs, elapsed time is going 
to be shorter when those faster engines can be used).

Some zIIP protocols leave the SRBs around for future use, so that it's 
less "startup/takedown" than "wakeup/put-to-sleep".

Peter Relson
z/OS Core Technology Design

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to