The statistics-for-engineers repository is an educational resource designed to teach IT operations engineers the statistical methods and techniques necessary for analyzing telemetry data in distributed systems. Created by Heinrich Hartmann and made possible through support from Circonus, a monitoring system with histogram support, this repository addresses the practical statistical knowledge that system operators need in their daily work. The primary language is Jupyter Notebook, making the content interactive and hands-on.
The repository's core purpose is to bridge the gap between raw monitoring data and actionable insights. Once organizations have set up monitoring systems and collected telemetry data, they face the challenge of extracting meaningful information to answer critical operational questions such as whether they are meeting their SLAs or how system performance has changed following updates. The tutorial covers foundational statistical concepts including probabilistic models, distribution summarization through mean values and quantiles, histogram analysis, and their mathematical relationships. It also addresses advanced topics like time series forecasting and scalability analysis, all presented with a focus on practical application rather than pure theory.
The instructional content is organized into discrete episodes covering specific topics. Episode 0 provides an introduction, Episode 1 focuses on visualizing data, Episode 2 covers histograms, Episode 3 addresses summary statistics, Episode 4 examines quantiles and outliers, Episode 5 explores forecasting, and Episode 6 delves into queuing theory. These episodes were originally presented at the 2015 Velocity Amsterdam conference and have been refined through subsequent presentations. The tutorial emphasizes hands-on knowledge using practical tools including UNIX command line utilities, gnuplot, and the iPython toolkit for handling, importing, analyzing, and visualizing telemetry data.
The repository has been presented at multiple major industry conferences spanning from 2015 to 2019, including SRECon events in Dublin, Düsseldorf, and Dublin again, as well as Velocity Amsterdam, StatsCraft in Tel-Aviv, Monitorama in Portland, and SRECon Dublin. The material has achieved significant academic recognition, with a writeup published in both ACM Queue volume 14 issue 1 and Communications of the ACM volume 59 number 7, demonstrating the relevance of the content to both practitioners and the broader computing community.
The repository includes practical datasets for hands-on learning, such as CPU utilization data from a machine cluster, database log files, API latency measurements, request rates across clustered nodes, and web latency measurements from multiple geographic locations. These datasets enable learners to apply the statistical techniques discussed in the episodes to real-world operational scenarios. The repository also provides a Docker-based bootstrap mechanism for setting up an interactive working environment, lowering the barrier to entry for engineers wanting to engage with the material.
GitGenius classification data indicates this repository is categorized across engineering, probability theory, quantitative analysis, and educational resource domains, with specific tags for statistical analysis, hypothesis testing, regression models, experimental design, and statistical inference. The repository shares overlapping contributors with ansible/awx, hashicorp/vault, and openhistogram/libcircllhist, connecting it to the broader infrastructure and monitoring ecosystem. The project maintains an active community through a mailing list for event notifications and uses the Twitter hashtag #StatsForEngineers to keep interested practitioners informed about upcoming workshops and presentations.
