* Understand why the latch or mutex is being held so long. Ask yourself, "Once the latch has been acquired, why is Oracle holding it so long?' Answer that question, and then ask yourself how you can influence the situation to reduce the time the latch must be held. As with the previous point, answers to this question are the focus of Chapters 6, 7, and 8.

* Determine multiple resolution strategies. Surprising to most DBAs is that once the preceding points have been addressed, you will actually be able to arrive at multiple solutions. And these solutions will not be guesses or even good guesses. They will absolutely impact performance for the better!

* Take appropriate action. This step deserves its own bullet point because usually one of the solutions involves changing a hidden instance parameter, which typically requires the Oracle instance to be cycled. I don't take this lightly. Most production system nowadays can't simply be cycled. You will need to prioritize the solutions and plan their implementation based on expected performance improvement and uptime requirements. Quantitative solution analysis is the focus of Chapter 9 and also OraPub's Advanced Oracle Performance Analysis class.