I don't know about zOS guests (so why am I wasting everybodies time?) We
have similar issues with TPF guests under VM and in LPARS... there is
*stuff* in the TPF CP (zstrc alter idletmr, which makes them aware of
VM/LPAR (with shared Lcp's (with dedicated cp's, it makes no difference));
a timer-pop mod sort of, but not enuf to pass test-idle...) probably not in
zOS... Anyways, measuring from within, doesn't make any sense and I believe
them; Measure from outside to size guest or LPAR CPU and storage...
ESAMON/MAP from Velocity-Software works great... DASD could be shared...
but If you're CPU bound... you could give a greater weight the z/OS native
LPAR (More or less, like CP SET SHARE and at the expense of the VM LPAR
(you don't get something for nothing, the 5lb bag is only 5lb's big), but
it's all about who's important and who's test... As Bill said, you'll save
some VM CP time running in an LPAR; and depending on your LPAR config,
you'll pick up a bit of physical management time and some LPAR management
time... worry not, LPAR is friend.
Gregg
office:404-322-2316 mobile:404455-1291 text page:[EMAIL PROTECTED]
"No plan survives execution" revised:26Aug05
file:\\Usfs01\Common\CPPS\VM\VMCapPlan.htm
William Boyer
<[EMAIL PROTECTED]>
Sent by: VM/ESA To
and z/VM [email protected]
Discussions cc
<[EMAIL PROTECTED]
.UARK.EDU> Subject
Performance of zOS guest.
10/13/2005 09:01
Please respond to
VM/ESA and z/VM
Discussions
<[EMAIL PROTECTED]
.UARK.EDU>
We were offered by an IBM Business Partner to have a performance study
conducted, but when they found we run the zOS under zVM they informed us
that IBM stated the tools would not work. Here is a copy of what was sent
to the business partner by IBM:
No, zOSPROD1 will not know about either zOSTEST or zOSPROD2. Not only that,
but RMF will not really know how much CPU it (itself) is using, nor will it
have any idea what the real configuration of the machine is. I'm looking
for the (now ancient) article explaining why you shouldn't trust MVS
performance data when it's running as a guest, but so far I haven't found
it.
BTW, the V=R and I/O Assist performance enhancements for high performance
guests are not available in an LPAR environment, so the instruction path
length is about double that of native z/OS performance. This has always
been true of guests under z/VM in an LPAR, and since all processors
starting with z/890 & z/990 will be LPAR only, z/VM is dropping support for
V=R and I/O assist even in a non-LPAR environment, starting with z/VM 5.1.
They must know this already, I'm sure.
This seems to indicate that if we would move our zOS to it’s own LPAR that
we could get better performance out of the zOS system.
QUESTION: How much better would the zOS perform in it’s own LPAR?
Currently we seem to be CPU constrained.
Can we still share the zOS DASD back to zVM in READ mode on zVM?
Note: The reason we run zOS under zVM is because in the old days we only
had parallel channels and we would have needed a CTC box to connect the two
LPARS. But now with a z800 and escon all I need is a cable. (Correct?)
Thanks in advance.
William L. Boyer
Senior IT Consultant / Senior Systems Programmer
Business and Technology Services
One West Pennsylvania Ave. ● Suite 700 ● Baltimore, MD 21204
Office: 410.832.8300 ext. 8419 ● Fax: 410.832.8329
www.emdeon.com/vips
This message is confidential, intended only for the named recipient(s) and
may contain information that is privileged or exempt from disclosure under
applicable law. If you are not the intended recipient(s), you are notified
that the dissemination, distribution, or copying of this message is
strictly prohibited. If you receive this message in error or are not the
named recipient(s), please notify the sender at either the fax address or
telephone number above and delete this message. Thank you.
