It is common to misunderstand what Oracle's wait interface is telling us. The information in Figure 2-8 is not referring to Oracle response time, end user response time, or what the users are experiencing. The top wait event's 53% value refers only to the wait time component of the Oracle response time; that is, what Oracle's kernel code instrumentation has provided.

Another important point is that our focus should be on the percentage of wait time and not the wait time itself. Referring to Figure 2-8, if I were to tell someone the time waited for scattered reads was 57 seconds and that this was a problem, they might laugh! First of all, Figure 2-8 is only a 30-second snapshot, so the wait time may not be large. And your colleague might expect a much larger scattered read time on his system, because his system is busier than your system. But regardless of these issues, looking at this system from an impact perspective and concentrating on what is affecting the sessions, it is clear that scattered read is the top wait event.

You may have noticed I have not mentioned the number of waits. This is because it is not an issue for our analysis. We are focusing on what affects our users, and that's time, not the number of waits. Look closely at the average wait times in Figure 2-8. You'll see that some wait events have an average wait time of less then 0.1 ms, while the largest average wait time is 110 ms. It is the combination of the number of waits and the individual wait time that affects our users, and therefore what's important. It is very common for the wait event with the largest number of waits to not be associated with the most time. Focus on the statistics that primarily impact the user experience. In this report, what impacts our users the most is the percentage of time waited.