Welcome to Performance change annotator, a tool for annotating time series data for changepoint analysis.

Please log in or register to get started.

General Introduction

Hello, This project is led by student Mohamed Bilel Besbes and supervised by Professor Diego Elias Damasceno Costa from REALISE Lab at Concordia University in Montréal, Canada, as well as:

• Suhaib Mujahid, Mozilla
• Gregory Mierzwinski, Mozilla
• Marco Castelluccio, Mozilla
• Alexander Serebrenik, Eindhoven University of Technology, Netherlands
• Philipp Leitner, University of Gothenburg, Sweden

Main Contribution

• General overview

  
  This project aims to optimize the performance engineering workflow at Mozilla, specifically the part of automatic detection by making it detect anomalies (regressions and improvements) faster and more precisely, thus saving investigation time for Mozilla performance engineers. The aim to do this by exploring alternatives to the existing method in use (Student T-test).

• Objective

  
  As we collected and explored the data from Mozilla systems, it was hard to have a baseline to compare Mozilla’s current Change Point Detection method (Student T-test). BUT, the issue is that the current data of the signatures timeseries contains backfills and retriggers, making it hard to replicate the original time series. Therefore, we wanted to make a dataset of experts-validated annotation to use it as a ground truth.

Your Contribution

As a Mozilla engineer, we provide this platform for you to annotate timeseries of signatures from Mozilla’s systems. You will have a tutorial below to familiarize yourself with the platform. Thank you for contributing to this!