Building High Performance Real-time Stream Processing Microservices
summary
Claims a real-time pipeline that combines streaming sources (Kafka/Pulsar) with in‑memory historical context using Hazelcast maps and map journal, runs deployable as a jar or on Hazelcast Viridian serverless, and performs trend detection via linear regression trend detection with thresholded alerts. Presents a benchmark of 1 billion tx/s on 45 nodes with ~30 ms latency and 99% accuracy. Shows code patterns: ingest to map, build trend and prediction maps, emit alerts to Kafka/websocket or files, and use auto‑scaling and eviction for production.
AI-written summary · about
chapters (10)
- 0:00Introduction and goals
- 4:49What is real time?
- 10:54Tools for stream processing
- 12:42Hazelcast platform overview
- 21:58Microservice architecture
- 25:28Scaling and deployment
- 27:38Demo setup and cloud options
- 34:48Real-time analytics with maps
- 40:57Summary and key takeaways
- 43:27Q&A session
related talks
06mPREVIEW: Real-time Stream Processing at Scale
22mBuilding FinTech Streaming Applications with Pulsar
46mComposable microservices for streaming analytics
36mStream All The Things!
30mEvent Streaming and the Future of Applications
51mEvent Streaming and Processing with Apache Pulsar
31mStreaming: Real-Time Architecture for Connected Products in Automotive
27mBuild ML Enhanced Event Streaming Apps with Java Microservices