While the other responders have made excellent suggestions/comments, I
suspect the problem (and "fix") may be simpler. This environment and
problem sounds like a good example of an eligible list being formed and that
one large Linux guest being stuck in it, not allowed to run. Doing a CP
IND Q will show the existance of an Eligible list (E3 value non-zero). If
this is the case, you can probably effecct an immediate improvement by
changing the SRM STORBUF values. Note that this MAY cause an increase in
the paging rates if the Linux guest is using all that memory, so you should
have a good paging subsystem to handle it. Although the average paging rate
may go up, it will probably be a more consistent paging rate rather than
the somewhat "bursty" pattern of a guest going in and out of the eligable
list. I'd try an initial STORBUF value something like:
CP SET SRM STORBUF 150% 200% 250%
and see if that makes a difference.
Of course, reducing the size (virtual storage) of the Linux guest would
accomplish much the same thing, but the STORBUF change is immediate and does
not require a stop and restart of the guest.
Mike Hammock
----- Original Message -----
From: "August Carideo" <[email protected]>
To: <[email protected]>
Sent: Wednesday, June 24, 2009 4:02 PM
Subject: Testing Linux Z/vm
We are preparing for a test and loading the database is taking forever.
Paging is high. I/O devices are less than 5% busy. CPU is never more than
20%. An Oracle file import is taking over 14 hours. Any Idea's with this
little bit of info to go on
Also is there a reason for the TCPIP machine to be set at 128M? the real
machine is 2 IFLs and 16G of storage. We see the paging coming from the
oracle machine( 2CPUs VM size =24G), Maint and TCPIP. Maint was set at
128M
and IPLing a non shared copy of CMS. We set Maint to 16M and IPL a shared
copy of CMS. His paging has stopped. I would like to set TCPIP at
something
less that 128M either 16M or 32M to stop his paging. Does this sound right
?
Augie
