I'm not smart enough to understand it! Why don't IBM just drop the price
for smaller customers?
On 25/07/2014 5:56 PM, Shane Ginnane wrote:
On Thu, 24 Jul 2014 22:00:16 -0500, Munif Sadek wrote:
Reason I am asking this question is that I can easily find out how long the
system has been capped and would like to predict
how long the system will be capped based on our projected workloads and service
requirements.
This is indeterminate as you don't know what the latent demand is for the
workload that caused the capping, and is still active. Capping causes the
workload(s) to smear horizontally (into the future) and impact on the workloads
entering the system later in time.
When I looked at this, a large spike in demand (prior to actual capping) can
impact the rolling average for several hours - especially if you are close to
the cap limit- and impact all future work. In one case, for an entire weekend.
I have tried to do manual calculation using 15 minutes interval MSU
calculation, my figures are very close to actual 4Hr rolling MSU but not
exact.
As the algorithms are AFAIK unpublished, "exact" is an unrealistic aim - be
happy you are close.
Capacity groups make this even more rubbery. I have found all LPARs in the group report
capping events even when the weights apparently should "protect" one/some from
run-away LPAR(s).
Looks good in theory, doesn't pan out so well under duress.
Shane ...
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
