Keynote – José Valim
summary
Announces two new libraries, gen_stage (≈1,200 lines) and flow (core ≈80 lines) that implement demand‑driven pipelines with backpressure, partitioning and windowing. Shows a real word‑count: eager Enum on a 2 gigabytes file never finishes (>10+ minutes), Stream finishes in 60s, Flow finishes in 36 seconds on a two‑core machine. Provides concrete APIs (producers/consumers/producer‑consumers, dispatchers, partition functions, dynamic supervisor) and working code and benchmarks, not vague promises.
AI-written summary · about
chapters (20)
- 0:00Platinum
- 0:13Lanyard
- 6:00roses are red\n violets are blue\n
- 8:30File.read("really large file")
- 16:03Flow stats
- 16:38Topics
- 18:56GenStage: Demand-driven
- 27:46max_demand: 10, min_demand: 5
- 30:49Dispatchers
- 31:23DemandDispatcher
- 32:26BroadcastDispatcher
- 32:45Partition Dispatcher
- 34:38GenStage goals
- 50:03Windows and triggers
- 52:00Flow (concurrent)
- 53:08Flow features
- 53:56Distributed?
- 57:34Library authors
- 58:35Inspirations
- 59:16The Team
related talks
02h 16mKeynote Day 2 - Full
35mKeynote | Stream Processing at LinkedIn: Then and Now
23mKeynote
47mAdvances in Stream Analytics: Apache Beam and Google Cloud Dataflow deep-dive
02mKeynote: Scaling Intelligence: Infrastructure as the Foundation
04mNeville Li Interview
11mKeynote: Open Source Stack Empowers the AI Era
19mKeynote: Building Bridges: Cloud Native and High Performance Computing