C
ChUIMonster
Guest
Of course I could be wrong but I'm pretty sure that you are barking up the wrong tree. Even if a miracle occurs and a previously unheard of secret feature not named "proshut" is revealed you still have the problem that the OS has also cached reads -- and you still need to flush those buffers. That's where the unmount/remount advice comes from. And that isn't going to go over well with a live database trying to keep those filesystems open. Your goal of determining if it is *likely* can be very easily satisfied by running ProTop. Simply look at the upper left hand portion of the dashboard (which comes up by default) . If your hypothesis is valid during "fast" performance you should see a smallish number of "os reads" and during poor performance you will see a much larger number. If that first order confirmation passes then drill down by looking at the specific tables and sessions (also shown by default) involved. If, on the other hand, "os reads" are not significantly different then you would just be wasting your time trying to prove it by flushing buffers etc.
Continue reading...
Continue reading...