Customer Reviews

Forecasting Oracle Performance

5 star:		(6)
4 star:		(0)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

This is not a book about Oracle Performance tuning. This is a book about Oracle Performance Forecasting. This should be evident after carefully reading the book.

Good books, like good professionals, are rare. This is why I have developed a tendency to choose carefully before buying one and, in the Oracle area in particular, I can smell the author's intellectual honesty and trustworthiness from the first page.

Forecasting Oracle Performance is one of those. I enjoyed the fluid style and closeness of the author. I enjoyed the simplicity and clarity, almost reminding me of Wittgenstein's famous quote ("everything that can be said can be said clearly"). I also enjoyed these short touches of humanism (cf Erlang Krarup's life).
On the forecasting subject, I liked the book construction and the quality of the examples.

Chapter 1 gives a landscape picture of Oracle forecasting and puts the reader in scope and context.
Chapter 2 introduces the reader to the basics of forecasting and the basic concepts over which the book is built: Transactions, response time, arrival time and basic formulas are described here.
Chapter 3 shows the limits of basic forecasting (essential formulas) and the problem of baseline and model selection. It shows how to increase forecasting precision with ErlangC or weighted averages. Most importantly, it shows why it is essential to understand the concepts and implications of the application of a given formula, model or method. The author is very careful in the choice of terms and always clears potential ambiguities. Those who know how difficult it can be to forecast will appreciate it.
Chapter 4 introduces to statistics applied to forecasting.
Chapter 5, on practical queuing theory, is probably the biggest chapter of the book. After a brief introduction to queuing theory, Little's law and Kendall notation, this chapter provides such a diverse set of examples (27!) making the topic very intuitive to non-specialists.
Chapters 6 & 7 describe forecasting methodology and workload characterization. The first describes the steps across a solid performance forecasting methodology, from the initial question to the actual forecast. The second deals with workload characterization: how to get system and Oracle data, how to choose the source and peak. It describes the workload modelling and the risks of data collection.
Chapters 8 and 9 are about models: ratio and linear regression models. They describe their respective foundations, limitations and advantages. Each chapter also contain several examples and case studies to illustrate the subject.
The last chapter deals with scalability models and their relationship with forecasting models.

Overall, I found this book very useful. You will find additional resources at the author's website (there is also a discussion forum on forecasting). The errata page is always up-to-date and some examples have even been extended. Great work!

A fantastic book. Not one of those that you'll be able to get through over a weekend but is one of those books that you constantly go back to for reference or validation. Some of the techniques take the mystery out of capacity planning and forecasting performance. Excellent.

Being an IT person I like to read such methodological books to refresh myself.
Before starting this book I had a feeling that it was little bit opposed to the secret of Stephen Hawking books where he had just used one famous formula E=mc^2. This book looked full of overwhelming mathematical formulas.
However, author presented material in a very easy to digest way. All formulas are absolutely necessary for understanding the Forecasting approaches.
These approaches are universal and don't depend on particular Oracle version. It's everything about applying proper forecasting models depending on the Forecasting question and workload data collected against complex Oracle environment. It's very straightforward if you read it fully.
It's MUST for those who are facing questions like" What would happen if the workload was doubled in three months? Does it make sense to buy 6 more CPUs to maintain agreed SLA? etc etc.
Book audience is
1) Oracle DBAs
2) Project managers/ Architects
3) Oracle developers

This book is quite unique on a market, because most of existing Oracle books describe Performance Troubleshooting.
Knowledge from this book will not become obsolete with new version of Oracle.

This is a great book to do FORECASTING, that is, if you already have a tunning problem you will not find the solution here. If you are in calm times regarding your application speed, this is when you have to do Performance Forecasting, for which this book is excelent. Very practical and just enough technicall to grasp the priciples.

I ordered 2 of these books at end of may, they arrived this week. Everything was ok even after a long trip to italy.
Thanks

I've always been a fan of Craig Shallahamer, an author who explains complex concepts in plain English.

This is not a trivial book, but it's an excellent book for performance tuning, a worthy book for the worthy scholar.

Oracle STATSPACK and the automated workload repository have provided a gold mine for the Oracle DBA to forecast future performance, and this is a great "get started" book to the statistical techniques that are required to perform performance forecasting.