Figure 9-7. Shown is a demonstration of the linear relationship between utilization and requirements. This graph is not based on theory or mathematical formulas, but data sampled from a CPU-intensive Oracle system. The solid line is based on the actual sample data, and the dotted line is a linear trend line. Their correlation coefficient is a very strong 0.9328.

Just as with CPU utilization, IO utilization can be calculated. However, because of the possibility of file system buffer caching and IO subsystem caching, our Oracle-focused utilization calculation is a worst-case scenario (no caching assumed), includes only the instances we sample data from, and does not include any non-Oracle-related IO. So, the worth of our calculation is limited (at best). The higher worth metric is Oracle's IO requirements, which we calculated in a previous section.

The IO team can use Oracle's IO requirements, apply whatever caching metric they wish, and also add any other IO-related metrics. While our utilization calculation has limited value, when comparing the theoretical worst-case utilization with the actual IO subsystem utilization, it will demonstrate the effectiveness of caching, changing IO subsystem capacity, and possibly various tuning efforts.