In a client/server architecture, there should be no time between the Oracle client process and the user's experience. However, with modern web architectures, this is not the case. Referring to Figure 5-13 again, the time between the Oracle client process and the user's web browser would be represented by unaccounted for time. This is your clue that the true performance issues reside outside Oracle's realm.

If the Oracle database is deemed guilty until proven innocent, I encourage you to either acquire or develop your own end-to-end response time monitoring system. This is not something you can throw together in a few hours. Indeed, there are companies based on providing infrastructure solutions6 to understand the user's experience. But it can be done using creative scripting, information from Oracle's performance views, and data from the networking group, and placing bots on a few strategically located PCs.

Session profiling has reached fanatical appeal. I spoke about this topic at a massive Oracle conference a few years ago. Afterward, a person meekly walked up to me. Looking over his shoulders, he quietly said, "Do you know what you just said? You are saying that profiling is not the best approach." He was clearly disturbed. Session profiling had become so personal to him it was like I was questioning this deeply held faith. My response was gentle, and I clearly restated my key point: profiling a session is not always the best approach. And, in fact, it is easy to be misled by reading more into the situation than is actually there. To bring clarity to this sometimes emotional topic, I will first define profiling and then present how to avoid being misled.