If you think about it, the weighted concept makes sense. Because the scattered read waits happen so much more often, the average IO read time should reflect this and be pulled toward the scattered read wait times. As Figure 9-21 shows, the weighted average value actually rounds to the average scattered read wait time of 0.093. While the difference may seem insignificant, not only can this have a dramatic effect when anticipating the impact of a performance solution, but it also makes the averages more realistic.

Our Oracle-focused solutions will concentrate on service time, queue time, or both. One solution to reduce the scattered read waits is to increase Oracle's buffer cache. There is plenty of memory, and (as shown shortly) there is also plenty of CPU available to handle the possible increase in cache management resources. Based on the size of the tables involved, a 1GB buffer cache should be able to cache the entire customer table.

Because the total queue time accounts for nearly 30% (28.9%) of the total response time, if queue time is eliminated, total response time could improve by as much as 30%. But there will likely be some other queue time, so to be conservative; let's say we anticipate a 20% decrease in queue time.