Clone this wiki locally
Gatling was born from the observation that the commonly used stress test tools don't have the required performance to stress nowadays applications. Moreover, any Java developer who used these tools was confronted with one of these situations:
- We can fire the tests only three times, because the license for LoadRunner is way too expensive...
- Oh my ! The structure of the pages has changed, we have to edit the whole script via the GUI, let's start clicking...
- Sigh! The customer chose to use OpenSTA, we do not have any choice but to use Windows VMs...
- JMeter cannot take that much load (strange, isn't it ?), we have to build a cluster of JMeters and monitor the application AND the injection nodes to be sure that the failure doesn't come from JMeter...
- The Grinder failed because of a memory leak, it's too hard to debug when you're not a Python developer...
Gatling is efficient
Gatling uses actors and async IO, so it's far more efficient that one-thread-per-user based solutions. It gives more accurate results with far less memory and CPU usage. What's the point in having to run the stress tool with more horse power than the system under test?!
Gatling scripts are elegant and concise
Simulation scripts should be written in a concise and user-friendly manner so that they can be handled by a version control system. Why is that? Because performing performance tests should be part of the release process, so a version of a scenario is to be bound to a version of the application sources, and one should be able to perform diffs between versions. This can't be achieved if the script are huge dumps of XML that can only be read from a specific GUI. Gatling provides a concise, elegant and easy to learn syntax:
scenario("Standard User") .exec( http("Access Github") get("http://github.com") ) .pause(2, 3) .exec( http("Search for 'gatling'") get("http://github.com/search") queryParam("q","gatling") ) .pause(2)
Gatling scripts are code
The standard DSL should be sufficient for most use cases. However, this DSL is actually written in Scala, on top of an open functional API. As a consequence, it can be easily extended with very few basic knowledge: no need to be a Scala hacker.
Gatling charts are meaningful
Gatling is free
Even if Gatling is not fully open source (due to dependencies licences), it is completely free of charge!