When reading this chapter, keep in mind everything that you've read about so far: Oracle's instrumentation, how to trace operating system processes, Oracle response-time analysis (ORTA), OraPub's 3-circle analysis methodology, that spinning on a latch is service time not queue time, and how to gather operating system utilization from v$osstat. If any of these topics seem unfamiliar, I respectfully ask you to turn back to those pages.

With the first four chapters as a foundation, we are now ready to begin completing our diagnostic focus.

In the section on CPU contention in Chapter 4, I introduced how to gather information about the operating system CPU consumption, utilization, and the CPU run queue. Here, we will focus on Oracle's CPU consumption. Obviously, Oracle's CPU consumption is a subset of all CPU consumed on the database server. For example, if the operating system CPU utilization is 65% with only a single Oracle instance on the box, it's a good guess that Oracle is consuming around 55% of all the CPU available on the database server. But in this section, we go a step further.