Interview and AMA with the creators of Stateful Functions
summary
Creators of Apache Flink’s Stateful Functions describe a runtime that exposes virtual actors with message passing and colocated durable state, using Flink checkpoints (typically stored on S3) to provide exactly-once semantics and a snapshot of the entire deployed system. The system multiplexes millions of logical instances onto a few physical workers, supports multiple SDKs, timers and async IO, and can call external databases at the cost of losing built-in consistency; not aimed at sub-millisecond bidding systems.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:13Why Stateful Functions
- 7:37Stateful Functions at this point
- 12:11Lessons from Akka
- 13:49Stateful Functions vs Akka
- 21:44Biggest misunderstandings
- 29:02Handling state outside
- 31:47Consistency
- 33:29Problems
related talks
55mStateful Functions: Polyglot Event-Driven Functions for Stateful Distributed Applications
24mStateful Functions: Flexible like a Microservice, Consistent like a Monolith
38mDeploying Stateful FaaS on Streaming Dataflows
27mStateful Functions 3.0: Towards a Cross-Platform Stateful Serverless Stack
43mStream Processing and Applications in the Modern Age
38mPolyglot Actor Model on Flink Stateful Functions
29mManaging State in Apache Flink
36mPaper Moon: A Comparison of Stateful Functions and Pulsar Functions