Durable Execution: The Next Leap for Microservices Architectures
summary
Durable execution tracks and replays function progress via a central server that proxies RPC over HTTP2 and stores a sequential event log so executions recover exactly to the precrash point, eliminating manual retry and reconciliation logic. The demo builds an e‑commerce flow with payments using stable idempotency tokens, delivery callbacks, Kafka-driven driver location updates and durable timers, showing retries after crashes that resume without double charges and succeed. Implementation is open source as a single-node Rust binary with a distributed and managed cloud release planned.
AI-written summary · about
chapters (14)
- 0:00Durable execution concept
- 4:01How durable execution works
- 8:46Durable execution demo
- 12:18Restate architecture
- 13:43Extending to communication
- 16:33State management
- 20:53E-commerce workflow demo
- 26:33Callbacks and messages
- 28:38Driver service state
- 31:35Durable execution benefits
- 32:42AWS Lambda integration
- 34:07Use cases
- 36:57Components summary
- 37:19Closing
related talks
44mThe Simplest Way to Build Resilient Applications
53mDurable Execution With Restate
45mThe simplest way to build resilient applications
27mThe simplest way to build resilient applications
48mFailure is not an option: durable execution + Dapr = 🚀
01h 00mReliable Agentic Systems need Durable Execution 🤖
52mMovin beyond request-reply: how smart APIs are different
49mManaging Resiliency in Microservices