The global data plane for autonomous driving from Day 0 to Day 2 — Stephan Reimann and Simon Putz
summary
Describes a global hybrid data plane: a distributed data fabric with multi-protocol support (files, objects, Kafka, POSIX/HDFS) fronting Kubernetes so analytics run in containers while storage stays persistent. Test fleets generate about 60 TB per car per 8 hours, sites stay independent with local VDI access, 3x replication and ~one‑month raw retention on edge clusters. Builds include CSI integration and unified analytics toolkits; initial deployment hit throughput issues traced to misconfiguration (threading) that is fixed and the system moves into stable Day‑2 operations.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:42Agenda
- 1:34Trends in Automotive
- 4:21The future of driving
- 9:55Managing data
- 13:07Data Fabric
- 15:13Challenges
- 22:52Questions
- 31:15Processing
- 33:55Realtime analysis
- 34:15Unified analytics
- 35:25Decentralized metadata
related talks
30mEnabling data-driven development for autonomous driving at BMW
30mData Platform Infrastructure at Rivian: A Deep Dive
06mData Driven Performance
40mBringing Automated Driving Software to the Production Vehicle
28mProgress For Big Data In Kubernetes
07mData Gravity and Kubernetes: Managing Large-Scale Data Ingest With Minimal Latency
24mCreating a Data Platform for Software-Defined Electric Vehicles
12mKeynote: Accelerating Autonomous Driving Network via Industry Collaboration