For the application analysis, referencing a top SQL report, you look for SQL consuming the most buffers (buffer gets, or what I call logical IO). You quickly identify three statements that are consuming around 90% of all the logical IO. As expected, the top two statements are part of the order-entry application. An application-focused solution is to tune the SQL with the objective of reducing logical IO. It's fine to also reduce the physical IO, but the Oracle response-time analysis and the operating system analysis clearly support first and foremost focusing on logical IO-that is, buffer gets.

In summary, the database server is clearly suffering from intense Oracle buffer cache management. This is supported by a CPU bottleneck, a very high level of easily identifiable logical IO SQL statements, and Oracle processes waiting to get a latch related to buffer cache management. Fortunately, there are a number of solutions that should resolve the problem, from an Oracle, an application, and an operating system perspective.

The 3-circle analysis method may seem strange and even uncomfortable at first, but once you do it a few times, you'll be amazed at how fast, thorough, and spot-on your diagnosis will be. Of course, you also need a good understanding of Oracle's architecture and its interaction with the operating system, which is what you'll get from this book.