IO testing - CALIBRATE_IO (11g) vs. other (10g)
Date: Wed, 3 Apr 2013 09:20:39 -0700
Message-ID: <CALgGkeD1P0TV+gbDReMRTr38KMdEhgmFYUueLWCANzMG_9f54g_at_mail.gmail.com>
Hi All,
We are strongly considering moving a key production database from 10.2.0.4 on Solaris to 11.2.0.3 on Linux (also, different server and SAN).
We've done a simple test of IO performance using DBMS_RESOURCE_MANAGER.CALIBRATE_IO on the new platform we (I) believe the result to be favorable.
I'd like to simulate DBMS_RESOURCE_MANAGER.CALIBRATE_IO testing on the current (10.2.0.4 production) system to try to get comparison metrics in order to set expectations on the new system.
Please note that the function DBMS_RESOURCE_MANAGER.CALIBRATE_IO is not available on 10.2.
I was thinking about using Orion or SLOB.
Looking over the web a little bit, I don't see anything about how DBMS_RESOURCE_MANAGER.CALIBRATE_IO actually works - i.e. how do I use Orion (or SLOB) such that it "mimics" CALIBRATE_IO?
The only issues I have using SLOB (on the production system) is that it requires:
a schema - that's probably OK as we have space for this,
a "small" SGA (to force physical IO) - we have a pretty large SGA on production (16GB); I can probably change it just for the test, however,
and I don't know if it will support Solaris [5.10].
Is there another tool which might be better for this?
Or, is this just not a good idea - that of "testing" in production (assuming we can do so with very little activity on the DB server - not down)?
I might be able to get some downtime to do this, however, that might prove difficult in this environment "just for a test".
TIA, Rich
-- http://www.freelists.org/webpage/oracle-lReceived on Wed Apr 03 2013 - 18:20:39 CEST