The Engineer's Guide to Streaming
summary
Compares Apache Storm, Flink and Apex by scalability, performance, durability and availability and claims those, not throughput numbers, decide platform fit. Notes Storm achieves single‑millisecond latency but recovers by resetting the entire topology so failures cascade. Describes Flink’s barrier checkpoints that enable fast recovery and configurable exactly‑once semantics. Presents Apex as production‑grade on YARN with a buffer server and per‑operator locality that lets systems dynamically reconfigure and avoid cascading failures, at the cost of higher complexity.
AI-written summary · about
chapters (8)
- 0:00Introduction to streaming
- 1:19Batch vs streaming systems
- 7:41Key differentiators in frameworks
- 9:26Apache Storm overview
- 16:30Apache Flink deep dive
- 27:05Apache Apex strengths
- 33:25State management in streaming
- 36:17Q&A session
related talks
31mHelp I need a stream processor - learning how to chose ...
28mSweet Streams Are Made Of These
37mStorm, a Distributed Real-time Data Processing System
42m"Distributed stream processing in Ruby"
38mFaster and Furiouser: Flink Drift
44mData Streaming: Which horse to bet on?
41mBeyond the Hype: When Apache Flink Solves Real Problems
39mEvaluating Streaming Framework Performance for a Large-Scale...