15 of 17 people found the following review helpful
Something New in the Realm of Oracle Optimization Books,
Verified Purchase(What's this?)
This review is from: The Art and Science of Oracle Performance Tuning (Paperback)
... This book is a must have for Oracle optimization specialists. Though geared mainly for the beginning-to-intermediate tuner, there is much useful information and food for thought to be found here for the more experienced specialist. The author stresses the importance of using wait events for optimizing, not the superannuated ratio-based approach.
This book helps the reader get away from bad optimization habits. Medical diagnosis is often used as a metaphor for the process, and Lawson often uses this (and other useful metaphors) in the book. Although he doesn't use the phrase, his book follows an old diagnostics adage: "When you hear hoof beats, think horses, not zebras." In other words, if you have latching problems, don't start pursuing what underscore parameters you can change, look to the prosaic root of the problem (almost inevitably to be found in bad design and bad SQL, and once in a while skimpy or dodgy hardware).
There is often a temptation to fling yourself at one technical problem after another on a system, hoping that the effort will magically solve the problem along the way. Lawson sets a step-by-step framework for working with the technical problem, the business situation, and most importantly, the people involved.
Many people in our industry are convinced that you can gather a few ratios and run a few scripts and find out everything you need to optimize a system. Occasionally this is true. But in the vast majority of cases you can gather at least as much information from the junior DBA you talk to at lunch about 'why the system is really like that' than you get from Statspack and your CD of scripts. More importantly, people can tell you what is important to their business, what plans they have for the system, and what areas are their strong and weak points. A lot of engineers (and optimization is a field dominated by the engineering mindset) have no problems setting up elaborate instrumentation on a system under study, gathering data and transforming that data into a good, objective report. That's the 'science' part of the work, and Lawson shows you a lot of fast, down-and-dirty scripts that can help you do just that. But he also devotes several chapters to the 'art' of optimization, a realm that is often alien to the engineering approach. Here the black and white answers obtained from studying the database and its various subsystems encounter the harsh reality of human frailty and financial sensibilities. This part of tuning requires people skills more than any other aspect of the RDBMS industry with the possible exception of sales. Tuning assignments are often a matter of reconciling the angry with the clueless, and improving a system in spite of the owners' best efforts to keep it in a hosed state.
This book is clearly based on years of experience and observation, and will be a good resource for years to come. There are a few anachronistic moments and places that could benefit from 'the latest toys', but overall the book is up-to-date (examples of the former are the use of 'UNRECOVERABLE' instead of 'NOLOGGING' in a piece of syntax and the stress on ordering table joins rather than cajoling the optimizer into doing the job better. The latter shows up in such lapses as having no mention of the incredibly useful v$SQL_PLAN table in 9i).
I heartily recommend this book.