The proper way to identify performance problems is to find out what it
is waiting for and identify bottlenecks. You identified only one
difference and so came to the conclusion 'It MUST be memory'. Did you
verify, apart from memory, this the only difference between the 2
systems?
Obviously, there is a reason for the 40% guideline. Apart from the
database itself, your client connections will also consume memory on
the server.
The more memory you are dedicating to Oracle's cache, the more likely
is it, the system is going to page fault like hell. You could of course
use continue to crank up memory, and wait for the system to page fault
like hell.
You would better try to *resolve* your problems, instead of *resort* to
panacea's like cranking up memory. Likely your problem is your
application, and sooner or later you will run into problems again.
--
Sybrand Bakker
Senior Oracle DBA
Received on Thu Sep 29 2005 - 08:36:51 CDT
