yes we are ficon, also using VDISk for swap, and we have changed the mach
size as suggested
and a few other things
we have Velocity monitoring S/W
we are waiting on the Philippines to retry the test
thanks all for the feed back,
Augie
As you can see right now nothing is going on - so nothing new to report
ind paging
No users in page wait
Ready; T=0.01/0.01 11:54:19
ind
AVGPROC-000% 02
XSTORE-000000/SEC MIGRATE-0000/SEC
MDC READS-000001/SEC WRITES-000001/SEC HIT RATIO-095%
PAGING-0/SEC STEAL-000%
Q0-00001(00000) DORMANT-00019
Q1-00001(00000) E1-00000(00000)
Q2-00000(00000) EXPAN-001 E2-00000(00000)
Q3-00001(00000) EXPAN-001 E3-00000(00000)
PROC 0000-000% IFL PROC 0001-000% IFL
LIMITED-00000
Ready; T=0.01/0.01 11:54:23
Tom Duerbusch
<duerbus...@stlou
iscity.com> To
Sent by: The IBM [email protected]
z/VM Operating cc
System
<[email protected] Subject
ARK.EDU> Re: Testing Linux Z/vm
06/26/2009 11:35
AM
Please respond to
The IBM z/VM
Operating System
<[email protected]
ARK.EDU>
With Oracle, someone is listing to the PC side of the house.
That is, inexpensive MIPS, inexpensive memory, expensive I/O. Hence you
want SGA to be as big as possible.
On the IFL side, MIPS are still relatively expensive, memory is still
relatively expensive, and assuming you are on FICON, I/Os are cheap.
Since you are just testing, cut the SGA in half, make the Linux guest about
512 MB bigger than the SGA, make the swap disks, VDISKS and prioritize
them.
You shouldn't have VM paging.
If you are planning on pounding I/O during the testing, don't forget I/O
balancing on you DASD subsystem. Make sure heavy I/O is spread across raid
arrays and controllers. The default tends to be accending CUA tend to be
on the same Raid adapter, until you jump to the next one.
Now, if you are really doing a full up Oracle test, that is, a full set of
applications, driven by some sets of test scripts, you are going to need
someone to configure your system for performance (of course you have a
performance monitor for VM....). Your current Oracle platforms were not
setup for performance in a day. Don't expect that on the mainframe, in a
day. But most testing, isn't in this arena and you can grow and tune the
system as you go.
It will be interesting to hear the progress of your testing.
Tom Duerbusch
THD Consulting
>>> Rob van der Heij <[email protected]> 6/25/2009 10:11 AM >>>
On Thu, Jun 25, 2009 at 5:56 AM, Marcy
> Your guest is bigger than your real storage. If you need to do that
overcommitting (and you may not - see Rich's question about SGA sizes),
Overcommitting resources is good, because it is the only way to
enforce sharing resources. But defining a single Linux guest larger
than your total z/VM storage is just bragging. And that is bad ;-)
As Rich says, you can't reduce the virtual machine size without also
adjusting Oracle's expectations (SGA and PGA). Otherwise Linux would
need to swap those areas and things go bad as well with two managers
involved.
Rob
--
Rob van der Heij
Velocity Software
http://www.velocitysoftware.com/
