Serious database administrators have a strong historical bias toward using UNIX-like systems for their servers. The first examples here are using Windows instead because the graphs produced are easier to read, and are therefore better for introducing the concepts of this section. But doing so reminded me why Windows is not the preferred database-hosting operating system for so many people.
Getting useful benchmark results requires the system be quiescent: free of other programs running that would spoil the results of what is intended to be measured. When booting into Windows Vista to generate these results, I discovered the TrustedInstaller process was hogging a considerable amount of CPU and disk resources. It turned out Windows Update had decided it was time to install the next major Vista Service Pack, it was downloading pieces in the background, and pushing me toward the upgrade at every opportunity. It was two hours later before I had completed all the background activity it compelled me to take care of and had an idle system capable of running these tests.