Anomaly Detection in Software Performance Testing

Having spent a fair amount of time developing software for a living, I've rarely ventured outside of a typical set of development techniques. That has changed a couple of years ago when I have picked up two data science courses on Coursera - Machine Larning by Andrew Ng from Stanford and Data Science by Roger Peng, Jeff Leek and Brian Caffo from John Hopkins University.

It helped while I was working on a project improving performance testing framework, as I've realized that anomaly detection techniques used for testing of complex industrial systems can also be used to test performance of software.

With that realization I could build analytics pipeline for the performance testing - defining the analysis question, getting and cleaning data, performing analysis and integrating the results back with the test tools.

Here's the demo of the approach that I've given at DataScienceLab meetup recently, showing integration of Cucumber, JMeter and R: https://github.com/cadmiumkitty/cucumber-jmeter-r-integration