Figure 2-18 is based on the OSM script rtsess9.sql, which focuses on a single Oracle session. The similar rtsess.sql script takes as its input a client identifier, which presents the possibility of profiling a session or a group of sessions.

If you are not confident that the end user is actively waiting for the application to respond, then while identifying a user or group of users using the client identifier solves the identification problem, the Net+Client time (based on the wait event SQL*Net message from client time) becomes useless. Again, the Oracle server process does not know if the wait is due to the network, a client process, or the user. Regardless of the reason, it will post the SQL*Net message from client wait event. A notable exception is when profiling batch processes. Since the Oracle client process never waits on an application user, we know all SQL*Net message from client wait time is associated with either the client process or the network. Even with this limitation, just as when profiling an entire Oracle system, we can still perform a very robust ORTA.

When someone is in pain but cannot verbally communicate, diagnosing the problem is difficult. Much like a very young patient, before Oracle instrumented its source code, it was limited in what it could say to us. Finally, Oracle instrumented its source code, and we began listening to it.