To complete an ORTA, we need a better understanding of where Oracle consumes CPU. For example, are the Oracle background processes consuming an unusual amount of CPU, or perhaps parsing is the culprit. Just as with wait event classification, Oracle allows a couple CPU-that is, service time-classifications. How we gather and classify Oracle CPU time and how that relates to the database server is what this section is about.

There are three approaches to gathering data about Oracle CPU consumption. The first is the traditional approach based on v$sysstat, the second is based on v$sys_time_model, and the final approach (which is covered in the final section in this chapter) is based on v$active_session_history. But before looking at the approaches, you should understand how Oracle perceives Oracle process CPU consumption.

Oracle is limited in its perception of Oracle process CPU consumption. While great strides have been made with the system time model, there are still two areas where Oracle can misrepresent CPU consumption. Keep in mind that these areas are of no consequence when the CPU subsystem is not massively bottlenecked.