8 There are many factors involved in deciding how fast redo log switches should occur. Paramount is recoverability and mean time to recover, not performance. Once the availability issues have been addressed, then work on slowing redo log switch time until the associated wait events are insignificant.

How many times have you been asked, "So what kind of performance improvement can we expect?" It's an honest and painfully practical question, which deserves an honest answer. Unfortunately, while Oracle professionals are proficient in many areas, one area where there is a glaring gap is in understanding the impact of their proposed performance solutions. The skill required to answer this question requires deep Oracle performance analysis proficiency.

This chapter's coverage borders on predictive performance analysis and some serious mathematics, yet I'll keep focused on simplicity, practicality, and value (for example, I will limit the number of Greek symbols and mathematical formulas to the absolute bare minimum). Furthermore, I will always tie the analysis back to the fundamentals: an Oracle response-time analysis, OraPub's 3-circle analysis, and solid Oracle internals knowledge. I do not intend to explain how to plan IT architectures. My goal is to provide substance, conviction, and useful information, and to motivate change toward scientifically ranking the many possible performance-enhancing solutions.