Forecasting Oracle Performance

by Craig Shallahamer

Overview

Now available in paperback-

What makes seasoned IT professionals run for cover? Answer: Forecasting Oracle Performance! Craig Shallahamer is an Oracle performance expert with over 18 years of experience. His book is the first to focus not on the problem of solving today's problem, but squarely on the problem of forecasting the future performance of an Oracle database. Other Oracle performance books are good for putting out fires; Craig's book helps you avoid all the heat in the first place.

If you're an IT practioner who appreciates application over mathematical proofs than you'll be pleasantly surprised! Each chapter is filled with examples to transform the theory, mathematics, and methods into something you can practically apply. Craig's goal is to teach you about real-word Oracle performance forecasting. Period. There is no hidden agenda.

This book is a kind of training course. After reading, studying, and practicing the material covered in this book, you to be able to confidently, responsibly, and professionally forecast performance and system capacity in a wide variety of real-life situations.

If you are more management-minded (or want to be), you will be delighted with the service level management focus. Forecasting makes good business sense because it maximizes the return on IT investment and minimizes unplanned down time. To those who think forecasting is a waste of money: well...obviously, they've never been on the evening news because their company lost millions of dollars in revenue and brand destruction because of poorly performing or unavailable systems.

Without a doubt, you will be equipped to deal with the realities of forecasting Oracle performance. But this book gives you more. Not only will you receive a technical and mathematical perspective, but also a communication, a presentation, and a management perspective. This is career building stuff and immensely satisfying!

Product Details

ISBN-13:	9781430242932
Publisher:	Apress
Publication date:	12/23/2011
Edition description:	1st ed.
Pages:	271
Product dimensions:	7.52(w) x 9.25(h) x 0.02(d)

About the Author

Craig Shallahamer has over 18 years of experience working in Oracle, empowering others to maximize their Oracle investment, efficiencies, and performance. In addition to being a consultant, researcher, writer, and keynote speaker at Oracle conferences, he is the designer and developer of OraPub's Advanced Reactive Performance Management and Forecasting Oracle Performance classes. He is also the architect of HoriZone, OraPub's service-level management product.

Table of Contents

About the Author     xiii
About the Technical Reviewers     xv
Introduction     xvii
Introduction to Performance Forecasting     1
Risk: A Four-Letter Word     2
Service-Level Management     3
Modeling: Making the Complex Simple     5
Model Types     6
Mathematical Models     6
Benchmark Models     7
Simulation Models     7
Differences Between Benchmarks and Simulations     8
Challenges in Forecasting Oracle Performance     9
Essential Performance Forecasting     13
The Computing System Is Alive     13
Transactions Are Units of Work     15
The Arrival Rate     15
The Transaction Processor     16
The Queue     17
Transaction Flow     18
The Response Time Curve     19
CPU and IO Subsystem Modeling     20
Method Is a Must     22
Data Collection     23
Essential Mathematics     25
The Formulas     25
The Application     27
What Management Needs to Know     29
Risk Mitigation Strategies     31
Tuning the Application and Oracle     31
Buying More CPU Capacity     32
Balancing Existing Workload     34
Summary     37
Increasing Forecast Precision     39
Forecasting Gotchas!     39
Model Selection     40
Questions to Ask     40
Fundamental Forecasting Models     42
Baseline Selection     46
Response Time Mathematics     47
Erlang C Forecasting Formulas     48
Contrasting Forecasting Formulas     57
Average Calculation     59
The Right Distribution Pattern     60
How to Average Diverse Values     61
Case Study: Highlight Company     65
Determine the Study Question     66
Gather and Characterize Workload     66
Select the Forecast Model     66
Forecast and Validate     67
What We Tell Management     72
Summary     73
Basic Forecasting Statistics     75
What Is Statistics?     75
Sample vs. Population     77
Describing Samples     77
Numerically Describing Samples     77
Visually Describing Data Samples      79
Fully Describing Sample Data     82
Making Inferences     89
Precision That Lies     91
Summary     93
Practical Queuing Theory     95
Queuing System Notation     95
Little's Law     99
Kendall's Notation     103
The Queuing Theory Workbook     106
Queuing Configurations and Response Time Curve Shifts     114
Observing the Effects of Different Queuing Configurations     114
Moving the Response Time Curve Around     119
Challenges in Queuing Theory Application     124
Summary     136
Methodically Forecasting Performance     139
The Need for a Method     139
The OraPub Forecasting Method     141
Determine the Study Question     141
Gather the Workload Data     144
Characterize the Workload     145
Develop and Use the Appropriate Model     145
Validate the Forecast     146
Forecast     151
Summary     151
Characterizing the Workload     153
The Challenge     153
Gathering the Workload     154
Gathering Operating System Data      155
Gathering Oracle Data     158
Defining Workload Components     161
Simple Workload Model     162
Single-Category Workload Model     163
Multiple-Category Workload Model     168
Selecting the Peak     181
Selecting a Single Sample     183
Summarizing Workload Samples     184
Summary     184
Ratio Modeling     185
What Is Ratio Modeling?     185
The Ratio Modeling Formula     186
Gathering and Characterizing the Workload     187
Deriving the Ratios     189
Deriving the Batch-to-CPU Ratio     189
Deriving the OLTP-to-CPU Ratio     192
Forecasting Using Ratio Modeling     194
Summary     198
Linear Regression Modeling     199
Avoiding Nonlinear Areas     199
Finding the Relationships     200
Determining a Linear Relationship     203
View the Raw Data     203
View the Raw Data Graph     205
View the Residual Data     206
View the Residual Data Graph     208
View the Regression Formula     211
View the Correlation Strength      212
If Everything Is OK, Forecast     213
Dealing with Outliers     214
Identifying Outliers     216
Determining When to Stop     219
Regression Analysis Case Studies     221
Summary     228
Scalability     229
The Relationship Between Physical CPUs and Effective CPUs     229
How Scalability Is Used in Forecasting     230
What's Involved in Scalability?     233
Speedup and Scaleup     235
Which Forecast Models Are Affected by Scalability?     236
Scalability Models     237
Amdahl Scaling     237
Geometric Scaling     240
Quadratic Scaling     240
Super-Serial Scaling     242
Methods to Determine Scalability     244
Physical-to-Effective CPU Data     244
Benchmark: Physical CPUs-to-Throughput Data     248
Real System: Load and Throughput Data     251
Summary     253
Index     255