I would agree that there is an 'underload' phenomenon when an LPAR is capped at 
a small percentage of the total available. 

Consider a LPAR capped at 10 in an 1000 MSU box, and assume there is nothing 
else running. A CPU intensive job kicks off, and the LPAR starts consuming all 
1000 MSU. It only takes a few sample cycles for the rolling four hour average 
to hit 10 and the cap activates. The higher the cap, the longer it takes for 
the R4A to rise, and the longer you run uncapped. 

Running over the limit does not bother me; that's covered in the sub capacity 
doc. But I run a 'penalty box' that is capped at a very low value to control 
software costs. 

My business need is to run LPAR1 at a R4A not to exceed x, and the group R4A 
not to exceed y.     

-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of 
Werner Kuehnel
Sent: Thursday, March 05, 2009 1:10 AM
To: [email protected]
Subject: Antwort: Re: Group Limit and Weights

Hal,
any answers you get are the theorie. The praxis looks a bit different. I 
made some analysis over the last weeks to understand how MSUs are 
distributed under capping. Sometimes it works like expected, but there are 
also times where 
- one lpar gets much more than expected, 
- the sum of group MSUs is significantly over the group limit,
- capping is on although the group limit isn't reached
- just one (of two) lpar is capped (what rules are then active?)
Probably this phenomenons are caused by the pretty small capacity each 
lpar has (40 MSUs and 12 MSUs on a 1202 MSU machine), so the granularity 
is too fine.

Werner Kuehnel


IBM Mainframe Discussion List <[email protected]> schrieb am 04.03.2009 
20:33:50:

> Any idea how quickly/frequently the reallocation occurs? 
> 
> Say L1 and L2 are group capped at 100, and are weighted 90 10 
> respectively. We are capped and L1 and L2 are running 50/50, but  L1
> abruptly needs more juice. How long would it take for L1 to get fed? 
> 
> -----Original Message-----
> From: IBM Mainframe Discussion List [mailto:[email protected]] On
> Behalf Of Al Sherkow
> Sent: Thursday, February 26, 2009 11:07 AM
> To: [email protected]
> Subject: Re: Group Limit and Weights
> 
> The capacity of the group is distributed to the LPARs in the group based 
on
> the normalized weights of the LPARs in that group. 
> 
> When the group limit is exceeded L1 would be 40 MSUs, (80% or 50) and L2
> would get 10 MSUs. If L1 does not require 40, then the excess can be 
used by
> L2 (just like without group limits). If L1 is not running at all then L2 
can
> have all 50 MSUs of the group. 
> 
> Al Sherkow, I/S Management Strategies, Ltd.
> Consulting Expertise on Capacity Planning, Performance Tuning,
> WLC, LPARs, IRD and LCS Software
> Seminars on IBM SW Pricing, LPARs, and IRD
> 
> 
> NOTICE: This electronic mail message and any files transmitted with 
> it are intended
> exclusively for the individual or entity to which it is addressed. 
> The message, 
> together with any attachment, may contain confidential and/or 
> privileged information.
> Any unauthorized review, use, printing, saving, copying, disclosure 
> or distribution 
> is strictly prohibited. If you have received this message in error, 
please 
> immediately advise the sender by reply email and delete all copies.
> 
> ----------------------------------------------------------------------
> 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

----------------------------------------------------------------------
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
NOTICE: This electronic mail message and any files transmitted with it are 
intended
exclusively for the individual or entity to which it is addressed. The message, 
together with any attachment, may contain confidential and/or privileged 
information.
Any unauthorized review, use, printing, saving, copying, disclosure or 
distribution 
is strictly prohibited. If you have received this message in error, please 
immediately advise the sender by reply email and delete all copies.

----------------------------------------------------------------------
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

Reply via email to