Slowdowndetection

Implementation of cluster data point evaluation and slow-down detection algorithm proposed for a recommendation system for performance testing decision making evaluated on CERN CMS uploader service. This requires an understanding of our entire approach which is present in the [article] (link to be updated). This approach is useful during performance regression testing of web service that deals with many test inputs.

Files information

• install.sh sets up python virtual environment and install all packages.
• dbscan_e9414f04.csv consists of clustered data based on DBSCAN clustering technique.
• selectdatapoints.csv data points selected randomly. This is done with random_data_points() method in slowdown_detection file.
• perfcibug.csv and pylintanalysis.csv files consists of two cases for evaluating our algorithm. These files are a result of sample profiling which we mention in our [article] ().

Instructions for running the slow-down detection algorithm.

• Run install.sh file
• Run slowdown_detection.py
• The output will be a json object which consists of decisions on different inputs about an update.