Event-Driven Infrastructure: Autonomous Operation and Unreliable Networks
summary
Claims an event-driven, state-machine architecture that lets agents make deterministic local decisions when controllers or networks fail; demonstrates a Rust-based runtime called Syspect/Sysinspect with durable network journaling so 100 queued commands survive master outage and resume, and with execution versus delivery semantics so clients verify outcomes locally. Shows tiny client that runs on Alpine with no Python, OpenTelemetry integration plus a Logjet relay, and an over-the-air model upgrade that propagates checksums to minions in about a second.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:07Observability costs
- 3:50Temporal logic basics
- 5:17Embedded vs server automation
- 8:03Autonomous operation
- 10:30Syspect verification
- 13:20Schrödinger automation
- 14:23Syspect workflow
- 16:52Open telemetry integration and log jet
- 19:19Live demo setup
- 27:20Minion registration
- 30:07Cluster upgrade demo
- 31:32Key takeaways and closing remarks
related talks
57mReliability Nirvana
29mRust: Embedded, Async, All the Way
40mProperty-testing async code in Rust to build reliable distributed systems
54mHow Events Are Reshaping Modern Systems
15mEvents are the Wrong Abstraction for Your AI Agents
44mDesigning for Autonomy
56mFrom Unreliable to Resilient: Mastering Distributed Systems
45mThe simplest way to build resilient applications