Everyone wants to release software faster, with fewer errors, so continuous integration is the watchword for modern release management
. But as Andrew C. Oliver writes in InfoWorld
, it’s surprising how few organizations integrate load testing early in the development cycle, since failing early is a lot less expensive than failing in production.
Continuous testing for the real world
One of the challenges of building a large, complex, and distributed application is creating a load testing environment that simulates real-world conditions. If your application is largely browser-based, tools like Selenium can generate specific tests that mimic real-world browsing at scale. You can use a continuous integration environment like Jenkins to run similar tests in a highly scalable and automated fashion.
Release management for continuous delivery often relies on automated use of relatively lightweight tools for test and performance monitoring. Oliver wades into controversy by dismissing the popular HP LoadRunner system for its cost and its heavyweight complexity. Read the comments section
and you’ll find passionate experts challenging Oliver’s assertions, supporting HP LoadRunner and talking about its suitability for performance testing of highly complex systems.
Whether you use simple testing tools or complex ones, the message is clear: Test early and test often, and make sure you incorporate load and performance tests into your continuous integration process. That’s the path to avoiding nasty — and expensive — surprises when you go into production.