Scaling live transactional data at the Tour de France
summary
Dimension Data captures per-second GPS from 192 riders over a 900MHz mesh that piggybacks TV microwave links, demuxes at the finish, and pushes raw feed to broadcasters while sending analytics to a cloud in Amsterdam. A rented data-truck proves fragile (storm moves the technical zone), forcing cloud migration; Akamai offloads ~98% and the site scales to ~120,000 concurrent users. The pipeline hits 99.5% data coverage and achieves 500ms analytics latency. Race Center and TV visuals run live; ML models (IBM Watson) deliver an 80% break-catch predictor and an effort index used by commentators.
AI-written summary · about
chapters (12)
- 0:00Introduction to Tour de France project
- 1:20ASO’s digital transformation goals
- 4:09Scaling challenges in the Tour
- 6:14Technical setup in mobile trucks
- 9:46Live data transmission solution
- 14:42Building the platform from scratch
- 16:48Telemetry data from bikes
- 21:19Data sources and analytics
- 27:02Machine learning for rider insights
- 31:20Visualizing race data effectively
- 35:12Predictive models for race outcomes
- 37:42Q&A session
related talks
20mInnovating at speed
06mData Driven Performance
29mUse sports data to elevate fan engagement and performance
40mFast Cars, Big Data - How Streaming Can Help Formula 1 #bbuzz
42mFast Cars, Big Data - How Streaming Can Help Formula 1
28mMoving Towards Real-Time OLAP - With First Principle Thinking and Guided by Access Patterns
28mTheory to practice: A guide to real-time architectures
38mCrunching data in real time for Fortune500 clients