Keynote | Real-Time Metrics at Fortnite Scale
summary
Metrics Royale extracts and aggregates arbitrary JSON fields with Apache Flink and Kafka, persisting partial two‑second aggregates into FoundationDB so multiple Flink clusters merge updates transactionally; it handles the Travis Scott Astronomical load (over 12 million concurrent players and ~200 million metrics/sec) and serves real‑time counts, sums and hyperloglog cardinalities to Grafana within seconds, using Jayway JSONPath for runtime metric definitions and forward‑rounding timestamps.
AI-written summary · about
chapters (26)
- 0:00Intro
- 0:25Our Agenda
- 1:06Our Team
- 1:28Fortnite
- 1:43What We're Facing
- 2:06Event Payloads
- 3:00What We're Solving
- 4:16Conquer a Use Case!
- 4:58Let's Strategize
- 6:40Original Version
- 6:55Shortcomings
- 8:04Goodbye, Old Friend
- 8:22What We're Looking For
- 9:06The Three Pillars
- 9:42Apache Kafka
- 10:23Apache Flink Very well maintained
- 11:31FoundationDB
- 14:30Flink's Power is the JVM
- 16:10Power of JsonPath
- 17:21User Powered Aggregations
- 19:05Persisting Metrics
- 19:47Merging Metrics
- 21:45Scanning Metrics
- 22:36Viewing Metrics
- 24:19Metrics For Everyone
- 25:05Lessons We Learned
related talks
10mInterview
01h 03mMonitoring Performance on Massive Games.
37mMeasuring 2.0 – Building the next generation webanalytics ...
20mHow We are Dealing with Metrics at Scale on GitLab.com
27mBatching Was Yesterday: Real-Time Tracking & Analysis For 100+ Million Visitors
49mKeynote: New Heights
52mReal-time-first metrics
13mLightning Talk: A Drop-in System To Accelerate Metrics Observability by 100x Us...