oracle

Oracle Performance Tuning 101 will change every
DBAs thinking about how to diagnose
and solve database performance problems in
a way which the DBA can understand what is really
going on behind the scenes. It is not about
black magic or hocus pocus Cache Hit Ratios
which do NOT explain the underlying behavior
of why the database is waiting and exactly
what is it waiting on. This Book helps you
understand how to find that wait information,
what it means, and some suggestions on how

to deal with it.

When I first read Chapter 2 of this book,
I remembered saying "WOW ... a new era is here
on how to track down bottlenecks in the Oracle
database, without scratching your head wondering why
the database wont go faster after we got
some hit ratio percentage back to 99%+".

As Gaja, Kirti and John have so eloquently
pointed out, tuning is mostly about finding what
is competing for the Oracle blocks. In the
end, this is causing the wait events and more
than likely all tracks lead back to expensive
logical I/O and underlying poorly written SQL.

For those that have been advocating spending tens
of thousands on new hardware or increasing the
shared pool ( which actually could cause adverse
behavior ) to fix performance issues, YOU need
to buy this book and read it. Then YOU need to
read it a second time as you would a college text.

Every Oracle DBA and every person who has
ever wrote a book on Oracle Performance Tuning,
should also read and adhere to the principles
of this book when attacking slowness in the
system/database.

Oracle Performance Tuning 101, Statspack and 10046
event data has helped me find and resolve every Oracle
Performance problem I have come across since I
changed my way of thinking on how to attack Oracle
performance problems.

A Fantastic and Bold Work !!!!! Great Job Gaja,
Kirti and John. Your efforts are to be highly
commended in the Oracle world.

An excellent book every Oracle shop should have. It is hard to find a book that at the same time is so rich in technical details, so easy to read, with so many cool and illustrative analogies (e.g. pctfree/pctused comparison to a restaurant), and with a fine human touch. The book has five parts plus Appendices. Part One established the groundwork for the remaining of the book. This is the first book that bases its approach and treats in full Oracle wait event tuning, as opposed to long traditional ratio-based tuning. Part Two deals with Application tuning: optimizers, statistics, access paths. Part Three is devoted to Instance and Database tuning: shared pool, buffer cache, redo log buffer tuning, as well as physical layout tuning: optimal block selection, avoiding fragmentation, partitioning. Part Four is called Specialized Tuning where parallel query and contention tuning are analyzed. Finally Part Five is devoted to "Environment Tuning", i.e. tuning of the OS (few flavors of unix and NT are separately treated) and, as far as I know, the most detailed coverage of RAID in any oracle book. The book is very easy to read, written very clearly and with a lot of humor. Every chapter starts with a related "myth" - a common knowledge that is either false, or not completely true, and ends with an "In a Nutshell" recap of the chapter. The book itself ends with a recap chapter, which in short reviews the whole tuning methodology (yet the RAID tables might have been omitted from this recap). The book has short queries that are illustrating the text, which are not difficult to retype, so the lack of a CD with the code from the book is not a disadvantage. Talking of queries, few more here and there would not hurt. I am not sure whether this is really a beginner ("101") book. Beginners will definitely find valuable stuff there, but for a full enjoyment in the book, and for getting the full benefit from it, I think that experience is very beneficial. A book is fun to read, and it is packed with technical details, interesting observations, and practical advises. It is the first book that fully deals with new emerging oracle tuning methodology: wait events based tuning. It is worth reading every page, and not just flipping through.

Two stars may be a little harsh; I'm sure the author is very knowledgable. However, there is a difference between making a lot of true statements and imparting knowledge to someone else (explaining).

First, be warned the book only is relevant to Oracle 7 and 8i. Oracle 9 tuning is not addressed. A lot has changed in Oracle 9 because of the automatic tuning features, so I feel this book is out of date and it is shameful that book sellers disguise this fact.

There is a lot wrong with this book from the standpoint of someone who needs to tune Oracle. If you are a full time DBA and spend a lot of time studying Oracle and Oracle is your life, then perhaps this is a useful book for your collection. In that respect, the "101" in the title is perhaps accurate - it opens the subjects that you will need to dig a lot deeper into in order to really get something useful done. However, if you need a guide to tuning your Oracle database, you probably will be lost and frustrated using this book.

The author in opening chapters lays out a tuning methodology which is basically "measure performance; locate bottleneck; tune appropriate component". Then he pretty much abandons that methodology and stuffs the book with brief descriptions of how Oracle does this and that, some related parameters, and some very general advice to wrap it up. But unless you take it much further yourself with other references and deep study, you will be hard pressed to know how to fix anything.

Many Oracle books I have come across suffer from the problem of being either 3 times too long or 1/3 as long as necessary. In other words, the author needs to choose a useful format: either simply provide an overview roadmap to more detailed information, or go the distance and provide a detailed enough amount of information to get the job done. This author similarly needs more or less to make this text useful.

I can't say the book is useless. Occasionally the author does give a brief formula or rule of thumb for sizing some parameter. But they are few and far between, and usually not in very important areas.

What most of us need is a "Tuning Guide". That is, a step-by-step methodology where measurements are taken and parameters are estimated based on the measurements for tuning the database. Iterative tuning may be required, but that's OK if it is layed out as such. As you work your way through the methodology, your Oracle instance and application come into "tune". I don't know if such a methology can be designed; experts may claim it requires "intuition" and "experience". If so, then don't bother writing a book; otherwise, it is the author's job to turn intuition and experience into a methodology that others can follow.

If a "Tuning Guide" is not the intent of the author but rather more deep understanding, then the author must follow the approach of building a crystal clear "model" of the system which identifies measurement parameters for estimating the state of a real system and identifies the "control parameters" which affect the performance of the system. Then the reader should be able to measure the system and perhaps deduce how to control the performance. This book falls far short of that goal.

Here is an example from the book that left me helpless:

"CAUTION: It is very counterproductive to Oracle system performance to over-allocate memory to one or more components of your shared pool. Over-allocation of memory here can and will cause significant parsing delays (in some cases we have noticed ten-minute response times for a query such as - select * from dual;)."

Then the author does not provide any real criteria as to when I might be straying into such a disasterous region. He goes on to talk about "free memory" for various shared pool area pools, and on careful study you might deduce that too much "free memory" could be a related problem, but then rather than give any formulas or hard advice, he covers his "bases" with the wishy-washy statement: "The key here is to manage the space appropriately and make use of all the available pools in your version of Oracle." I'd love to - tell me how!

I won't dwell on his erratic writing style which frequently tosses in chirpy lingo such as the subheading: "Hey, Oracle - What Is Your Plan of Action (P.O.A.)?". His use of analogies is weak and half-hearted such as his analogy for resource contention with "children all wanting the same toy".

Basically, you know when you have a killer book that is a great tool in your toolkit. This one ain't it.