Book Image

Master Apache JMeter - From Load Testing to DevOps

By : Antonio Gomes Rodrigues, Bruno Demion (Milamber), Philippe Mouawad
Book Image

Master Apache JMeter - From Load Testing to DevOps

By: Antonio Gomes Rodrigues, Bruno Demion (Milamber), Philippe Mouawad

Overview of this book

Load tests help identify the maximum number of requests a software system can handle. One popular open source tool for load testing is JMeter. By leveraging the features and capabilities of JMeter, you can perform extensive load testing and fix issues in your application before they become problematic. This book is written by JMeter developers and begins by discussing the whole process, including recording a script, setting it up, and launching it, enabling you to almost immediately start load testing. You'll learn the best practices that you must follow while designing test cases. You'll also explore the different protocols offered by JMeter through various real-world examples. Finally, you'll see how to integrate JMeter into the DevOps approach and create professional reports. You'll discover ways to use the eco-system of JMeter to integrate new protocols, enrich its monitoring, and leverage its power through the use of the cloud. By the end of this book, you'll know all that's needed to perform comprehensive load testing on your applications by using all the best practices and features of JMeter.
Table of Contents (14 chapters)

Some Tips to Present the Results

After a load test, we will have found a lot of things. It's then time to present them to:

  • Give an idea of response times and stability before going into production
  • Allow Ops to fix configuration problems
  • Allow developers to fix problems
  • Ask for budget increases for hardware/licenses
  • Inform the CTO and so on

But presenting numbers through a graph is not as easy as it seems.

To help us, let's look at some tips.

Don't Average Percentiles

Computing the average of percentiles is mathematically incorrect.

Be careful with the tool used to display results:

  • Check whether it can compute the new percentile (see the Prefer Percentiles over Averages section for more details)
  • Check how the tool proceeds with metrics if the screen resolution is too small to display all points

For more details, see Gil Tene's blog post called "You can't average percentiles. Period." (http://latencytipoftheday.blogspot.com/2014...