Section 1.1. You re Doing It Wrong

   

1.1 "You're Doing It Wrong"

One of my hobbies involves building rather largish things out of wood. This hobby involves the use of heavy machines that, given the choice, would prefer to eat my fingers instead of a piece of five- quarters American Black Walnut. One of the most fun things about the hobby for me is to read about a new technique that improves accuracy and saves time, while dramatically reducing my personal risk of accidental death and dismemberment. For me, getting the "D'oh, I'm doing it wrong!" sensation is a pleasurable thing, because it means that I'm on the brink of learning something that will make my life noticeably better. The net effect of such events on my emotional well-being is overwhelmingly positive. Although I'm of course a little disappointed every time I acquire more proof that I'm not omniscient, I'm overjoyed at the notion that soon I'll be better.

It is in the spirit of this story that I submit for your consideration the following hypothesis:

If you find that Oracle performance tuning is really difficult, then chances are excellent that you're doing it wrong.

Now, here's the scary part:

You're doing it wrong because you've been taught to do it that way.

This is my gauntlet. I believe that most of the Oracle tuning methods either implied or taught since the 1980s are fundamentally flawed. My motivation for writing this book is to share with you the research that has convinced me that there's a vastly better way.

Let's begin with a synopsis of the "method" that you're probably using today. A method is supposed to be a deterministic sequence of steps. One of the first things you might notice in the literature available today is the striking absence of actual method. Most authors focus far more attention on tips and techniques than on methods. The result is a massive battery of "things you might want to do" with virtually no structure present to tell you whether or when it's appropriate to do each. If you browse google.com hits on the string "Oracle performance method," you'll see what I mean.

Most of the Oracle performance improvement methods prescribed today can be summarized as the sequence of steps described in Method C (the conventional trial-and-error approach). If you have a difficult time with Oracle performance optimization, the reason may dawn on you as you review Method C. One of the few things that this method actually optimizes is the flow of revenue to performance specialists who take a long time to solve performance problems.

Method C: The Trial-and-Error Method That Dominates the Oracle Performance Tuning Culture Today

  1. Hypothesize that some performance metric x has an unacceptable value.

  2. Try things with the intent of improving x . Undo any attempt that makes performance noticeably worse .

  3. If users do not perceive a satisfactory response time improvement, then go to step 1.

  4. If the performance improvement is satisfactory, then go to step 1 anyway, because it may be possible to produce other performance improvements if you just keep searching.

This trial-and-error approach is, of course, not the only performance improvement method in town. The YAPP Method first described by Anjo Kolk and Shari Yamaguchi in the 1990s [Kolk et al. (1999)] was probably the first to rise above the inauspicious domain of tips and techniques to result in a truly usable deterministic sequence of steps. YAPP truly revolutionized the process of performance problem diagnosis, and it serves as one of the principal inspirations for this text.


   
Top


Optimizing Oracle Performance
Optimizing Oracle Performance
ISBN: 059600527X
EAN: 2147483647
Year: 2002
Pages: 102

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net