Let's say you're an Oracle tuning hotshot. You perform a stellar OraPub 3-circle response time-based performance analysis, and reduce the Oracle server process time by 80%, from 10 ms to 2 ms. That would be an amazing feat to be sure. But think about the impact from the end user's perspective. What used to take 50 ms, now takes 42 ms. So the user experiences a 16% improvement. When the users are asked how the system is working now, they might say it feels a little faster. So, doing a fantastic database-centric tuning job did not make enough of a difference to satisfy the end user. But the situation can become even more bleak.

Suppose the end-to-end response time is 115 ms, using the architecture shown in Figure 5-13, and the timing breakdown as follows:

Let's say you've a true gift from the Oracle performance gods and reduce Oracle server time by 86%, from 7 ms down to a mind-blowing 1 ms. Even so, the users' experience will improve only from 115 ms down to 109 ms. When they don't feel the 5% end-to-end response time improvement, your achievement will not get the appreciation you think it deserves. The point is that as DBAs, our impact, while perhaps technically miraculous, may have little effect on the end-user experience.