We have a Dell R710 running Oracle Standard Edition Two on two xeon X5687 processors and 192gb of dual rank (12x16gb) ddr3 running at 1333MHz. This is a standard database running on Solaris 10, supporting a website using a similar set up to LAMP, with the exception that, rather than using MySQL, or using Oracle 11 G. We measure responsiveness of the whole platform using a service called Monitis. It times the process of going to the website hitting the back end database and then returning results to the web viewer.
Our problem is that, after we run a job like datapump or do a large insert, the performance of the machine deteriorates noticeably and stays on that new poorer performance track for hours and days at a time UNTIL the next system reboot.
We've noticed that the performance can be rectified immediately on a full system restart. On Wednesday night, for example, the system which had been running very slowly snapped back to great performance metrics after a reboot. It stayed this way until the next evening at 10 PM when the nightly data pump job was run. Responsiveness crashed dramatically at the outset of the process and, while it improved a lot once the process was finished, it has established a new lower plateau of performance that has persisted over the last 24 hours.
We've ruled out any kind of networking issues and are positive that the problem resides in the Oracle database server.
Is there an issue with memory leaking or too much ram for Oracle 11g? This problem kicked in on Feb 11 when we 1) brought system ram up to 192 gb and 2) swapped out Xeon E5520's for X5687's.