Complex Event Processing on Arbitrary Data in Real-Time
summary
Myrmas is a low‑level CEP kernel that converts semi‑structured logs into term‑to‑triples flattening (uuid, 16‑bit per‑level element ids, scalar/container value) so patterns unify cheaply with a custom Rete‑style engine and bloom‑filter prefilters, limiting data movement. Element ids use 16 bits per depth with 3‑bit tag + 13‑bit position (≈8,000 slots per node). Mixed Erlang plus embedded Milton SML runs hot paths via a thin C shim, ships as the open source core for AutonomicI/O, and materially reduces re‑execution while enabling distributed matching.
AI-written summary · about
chapters (12)
- 0:00Introduction and Overview
- 1:01Engineering Goals
- 2:45Action and Examples
- 4:26What We Don't Do
- 6:03Data Structuring
- 8:07Pattern Recognition System
- 9:18Data Representation
- 12:29Pattern Matching Process
- 13:42Pattern Description Language
- 19:37Miramus Kernel
- 20:31Embedded Optimization Language
- 24:19Q&A and Conclusion
related talks
53mQuerying Events
34mComplex Event Processing Made Easy with Cassandra and Quine Streaming Graph
50mProcessing Billions of Events in Real-Time with Heron
51mThe Erlang VM or How I Started Loving Parallel Progamming
46mEvents Analysis with Riak, Perl and Erlang at Booking.com
01h 00mHyper Speed: When Big Data Blooms
54mMicro-Mind Services: Building Complex Behavioral Systems
38mPostmortem: Evolution of an Event recording solution