Licensing with VWLC can save a bundle on z/OS and other software charges
unless you are regularly running your box at 100% for four hours
straight every month. It saved us over $10K per month when we started
using it; but it also changes the game from one of "getting what you
paid for" to one of "paying for what you get" (or more accurately,
paying for your peak usage). If someone introduces a program bug that
eats CPU and it runs all night before it's caught and fixed and that
turns out to be your monthly peak four hours, that bug may cost a
substantial amount of real money in software costs, and management
doesn't tend to like budgetary surprises, especially one that is
unnecessary.
Monthly VWLC charges are based on the peak four hour combined average of
all LPARs running that software during a given month, so constraining
one LPAR doesn't cut it. To be able to tell management our software
costs won't exceed "x" for this month, you need to cap the combined
usage of all LPARs running the VWLC software, but you still want a
single LPAR to get up to the max if the other LPARs don't need it.
Software costs these days tend to be much more than hardware costs, and
hardware upgrade increments can be much larger than the immediate need.
If you operate under a tight budget, you may need equally tight
control over when your VWLC costs are allowed to rise on a machine that
is normally underutilized.
Joel Wolpert wrote:
The problem with "Defined Capacity" is that it is done at an lpar level.
There is still no way to guarantee that the machine as a whole will not use
more than 28 MSU's.
Why do you want to cap the machine. It makes no sense from a capacity
standpoint. Don't you want to get what you paid for?
Joel Wolpert
Director - Performance and Capacity Planning
Shared Data Center
Securities Industry Automation Corporation
2 Metrotech Center
New York, NY 11201
(212) 383-3323
[EMAIL PROTECTED]
>
Look at 'Defined Capacity'. This allows you to set a MSU cap at 4 hour
rolling average per LPAR. The 4 hour average will allow you to exceed
the cap as long as the 4HrA does not exceed the cap. This is good if
your workload has peak demand beyond 28 MSU but only for a short time
(LT 4 Hr). I have heard/read somewhere (possibly what Eric refers to)
that due to rounding/reporting you could consistently achieve a little
more than the cap (say up to +1) with no ill effects. If you are using
VWLC then IBM will not charge more than defined capacity even if it is
exceeded for some reason.
The only other way you could cap would to define and activate a third
LPAR (no OS needed) and hard cap all LPARs so your 2 'real' LPARs have
28 MSU and the 'dummy' third has 4 MSU.
Unfortunately I know of no way to 'share' a cap between 2+ LPARs (but
really really wish I could).
--
Joel C. Ewing, Fort Smith, AR [EMAIL PROTECTED]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
