Running Machine Learning Analytics On Traces
summary
Proposes storing logs and traces in cloud object storage and feeding them into an open‑source Hazelcast streaming layer that keeps hot state in memory, runs models close to data, and auto‑scales. Demonstrates running Python ML on the JVM, using linear regression as an example, routing alerts to Slack/webhooks, and handling 1 billion events on 45 nodes in a benchmark. Offers deployment options: local, edge sidecar, or separate model farm and a switchable ensemble approach to avoid downtime.
AI-written summary · about
chapters (7)
- 0:00Introduction to real-time ML analytics
- 4:20Why traces need machine learning
- 8:53Introducing Hazelcast platform
- 14:56Infrastructure and cloud deployment
- 20:36Scaling machine learning models
- 25:41Demo and practical implementation
- 30:24Q&A and closing
related talks
27mBeyond Logs: Real Time Anomaly Detection with Machine Learning
43mBeyond Logs: Real-Time Anomaly Detection with Machine Learning
24mData analytics with distributed tracing data
10mML-based Log Analysis for Faster Debugging
37mBig Data Analytics, Machine Learning, Real Time Processing
14mPerformance Optimization in High-Volume Transaction Systems
01h 21mBuilding a ML Model Observability pipeline
41mInfrastructure for ML Applications