Polyglot Actor Model on Flink Stateful Functions
summary
Compares classic actor model (lightweight processes, cheap millions of actors, low-latency distribution) with Flink Stateful Functions that store durable per-key state in RocksDB backend, provide exactly-once processing, and treat functions as virtual actors swapped to disk. Shows word-count example (one counter per word), claims terabytes of Flink state and millions of events are practical, demonstrates a polyglot pipeline (polyglot (Python/Scala)) used at Bluechip for direct-mail scoring, and warns blocking HTTP required Kafka RPC workaround.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:31What is the actor model?
- 1:55Example - Tail one file
- 3:31Example - Tail many files
- 5:38Example - Slack
- 10:01What is Statefun?
- 12:49Example - Wordcount
- 15:19Functions are virtual
- 24:19Direct Mail
- 26:51Fraud (Planned)
related talks
55mStateful Functions: Polyglot Event-Driven Functions for Stateful Distributed Applications
42mInterview and AMA with the creators of Stateful Functions
34mHands-on Stateful Serverless Applications with Kubernetes and Stateful Functions
38mDeploying Stateful FaaS on Streaming Dataflows
43mStream Processing and Applications in the Modern Age
27mStateful Functions 3.0: Towards a Cross-Platform Stateful Serverless Stack
37mBuilding and operating large-scale streaming applications at King
29mManaging State in Apache Flink