Optimizing Java

Chapter 13. Profiling

The term profiling has a somewhat loose usage among programmers. There are in fact several different approaches to profiling that are possible, of which the two most common are:

• Execution

• Allocation

In this chapter, we will cover both of these topics. Our initial focus will be on execution profiling, and we will use this subject to introduce the tools that are available to profile applications. Later in the chapter we will introduce memory profiling and see how the various tools provide this capability.

One of the key themes that we will explore is just how important it is for Java developers and performance engineers to understand the way in which profilers in general operate. Profilers are very capable of misrepresenting application behavior and exhibiting noticeable biases.

Execution profiling is one of the areas of performance analysis where these biases come to the fore. The cautious performance engineer will be aware of this possibility and will compensate for it in various ways, including profiling with multiple tools in order to understand what’s really going on.

It is equally important for engineers to address their own cognitive biases, and to not go looking for the performance behavior that they expect. The antipatterns and cognitive traps that we met in Chapter 4 are a good place to start when training ourselves to avoid these problems.

Introduction to Profiling

In general, JVM profiling and monitoring tools operate by using some low-level ...