Paul is one of the few.
Yes moving memory and physical devices can take some time (as opposed to
resources), mainly due to the processes involved in moving them. Think
about each step the system must take and it'll make sense, but it does not
take significant resources that I've found in monitoring several hundred
systems.
Just moving CPU back and forth, does not require much on a system that's not
really busy in total. If you have multiple partitions all calling for more
CPU, you'll see a pause while the system sorts out who gets what.
Also: The shared processor mode "Weight" is NOT A PRIORITY!!! It is very
different than priority and take the time to understand it before you set it
at too low or too high a value. Most customers that have trouble with
moving CPU have fallen into this trap. I even know some Premier Partners in
Florida whose technicians set this wrong every time I see a system they set
up. Be careful whoever sets up your partitions (or helps you) understands
the settings in there.
--
Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects
-----Original Message-----
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of
Steinmetz, Paul
Sent: Monday, February 16, 2015 9:39 AM
To: 'Midrange Systems Technical Discussion'
Subject: RE: PowerVM overhead
I have found that there can be substantial overhead involved in moving
dynamic resources, especially memory.
Because of this, I opt to run CAPPED LPARs, cpu and memory allocated.
Paul
-----Original Message-----
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of
Mitchell, Dana
Sent: Monday, February 16, 2015 10:31 AM
To: Midrange Systems Technical Discussion
Subject: PowerVM overhead
Is there measureable overhead assoiciated with PowerVM hypervisor having to
frequently allow an LPAR to exceed its CPU allotment and draw from unused
capacity in the shared CPU pool? We have an 8202-E4D with 2 LPARs. One is
allocated 0.75 of a core, the other is 0.25 with 2 reserved, the Shared
pool contains 3 cores. The larger of the two lpars pretty much runs over
.75 of a core throughout prime shift. It has been proposed that we work to
allocate CPU closer to the actual usage to reduce as much as possible the
situation where the hypervisor has to tap into unused capacity from the
reserved amount and/or the other LPAR. Does anyone have any evidence one
way or the other on this quandary?
Thanks
Dana
Attention: This electronic document and associated attachments (if any) may
contain confidential information of the sender (SHAZAM Network) and is
intended solely for use by the addressee(s). Review by unintended
individuals is prohibited. If you are not the intended recipient: (i) do not
read, transmit, copy, disclose, store, or utilize this communication in any
manner; (ii) please reply to the sender immediately, state that you received
it in error and permanently delete this message and any attachment(s) from
your computer and destroy the material in its entirety if in hard copy
format. If you are the intended recipient, please use discretion in any
email reply to ensure that you do not send confidential information as we
cannot secure it through this medium. By responding to us through internet
e-mail, you agree to hold SHAZAM, Inc. and all affiliated companies harmless
for any unintentional dissemination of information contained in your
message. Thank you.
--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe, unsubscribe,
or change list options,
visit:
http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take a
moment to review the archives at
http://archive.midrange.com/midrange-l.
--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe, unsubscribe,
or change list options,
visit:
http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take a
moment to review the archives at
http://archive.midrange.com/midrange-l.
midrange.com
