Absolutely agree!  I've done various tech sessions, and customer
presentations on this for years.
As long as you know what happens when your system runs at 100%, it's a good
thing.  There are NO
roll over minutes. What you didn't use a minute ago, is not available in the
next.  If important work
is meeting its goal, why not get your money's worth?

-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[email protected]] On
Behalf Of Timothy Sipples
Sent: Thursday, September 18, 2014 9:21 PM
To: [email protected]
Subject: Re: High CPU Utilized

As other posters have mentioned, and as I'll echo and agree, it's perfectly
normal for CPU-bound batch to drive your z/OS system to 100% CPU utilization
for some period of time. As long as you are meeting your service level
agreements (SLAs), no problem. The system is doing exactly what it's
supposed to do: allocate resources to execute your batch according to the
service goals you've set while concurrently running other programs (such as
online transactions) according to their service goals.
Many, many customers are "batch peak" customers -- it's thoroughly common.

If you'd like to reduce your CPU consumption and thus (probably) lengthen
the amount of time your CPU-bound batch job(s) take to run, you can do that
-- if your business users find that arrangement acceptable. One way is to
define a "softcap" either on an individual LPAR basis or across a group of
LPARs. Note that a "softcap" allows the CPU utilization to spike above the
cap but will definitively cap the 4 hour rolling average (4HRA).

But there's absolutely nothing inherently wrong with 100% CPU utilization in
z/OS. Often that's a good thing, actually, especially if you're meeting all
your SLAs (including elapsed batch execution time) with just a "little bit"
of margin to spare. In that perfect case it means you're getting your
money's worth -- you've got exactly the right amount of capacity to meet the
SLAs for the workloads you need to run.

I don't recommend trying to second guess z/OS. Configure z/OS --
particularly WLM -- with the service goals you need according to your SLAs,
and let z/OS figure out how to get there. Throw the batch jobs into the mix
when they're ready to run, not later. Then control the overall utilization
if you wish, particularly with one or more softcaps. If the critical online
transactions go quiet when your call center loses power (for example), z/OS
will run more batch (for example) at that moment in time -- and that's a
very good thing. Let z/OS be z/OS and do its job(s) which it does very, very
well.

----------------------------------------------------------------------------
----------------------------
Timothy Sipples
IT Architect Executive, zEnterprise Industry Solutions, AP/GCG/MEA
E-Mail: [email protected]
----------------------------------------------------------------------
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

Reply via email to