From an operating system perspective, because Oracle processes are waiting for blocks outside Oracle's buffer cache and on average take 22 ms8 to be received from the IO subsystem, you can clearly expect an IO bottleneck. You have not implied that there is a problem with the IO subsystem or that the SQL is poorly written, but the fact remains that it takes 22 ms to complete a multiple-block IO call.

After using standard operating system commands to find the CPU, IO, memory, or network bottleneck (this will be discussed in Chapter 4), sure enough, you discover there is a hot disk array. And when you dig a little deeper, you notice the tables are involved in multiblock reads that are located on the hot disk array! Is it possible to reduce the IO subsystem multiblock read times? That is what you will talk about with the IO subsystem team (which includes storage and capacity management personnel). But regardless of the IO subsystem team's cooperation, the operating system link to both the Oracle and the application subsystems has clearly been established.

At this point, you have a three-way confirmed relationship, clearly showing issues and opportunities for performance improvement. This may sound overly simplistic, and in some ways, it is. However, the method, the process, and the steps taken are essentially the same, regardless of the Oracle server configuration or complexity.