Workflows, a new abstraction for distributed systems
summary
Claims a new abstraction gives distributed systems exactly-once semantics by turning volatile functions into durable executions using event-history replay and idempotent activities. Implements this in production platforms (Temporal, Cadence, Azure Durable Functions, Infinitic) using coroutines, logs, and task retries. Shows code-level mechanism: yield closures, record task results, replay to checkpoints. Reports real deployments at Netflix Snap Uber (and Box) and multi-language support (Go Java JavaScript TypeScript PHP with Python and Rust coming).
AI-written summary · about
chapters (5)
- 0:00Start
- 0:52What Is the Difference between a Proof of Concept and a Production System
- 2:06Volatile Executions
- 26:17Retrying Tasks
- 28:08Rehydrating State
related talks
45mDapr Workflows: How to Write Reliable, Fault-Tolerant Applications...
25mDapr Workflows: The Key to Simplifying Business Process Orchestration
01h 00mReliable Agentic Systems need Durable Execution 🤖
36mOnyx: Distributed Workflows for Dynamic Systems
31mBuilding Asynchronous SOA for Modern Apps
44mThe Simplest Way to Build Resilient Applications
19mDistributed Workflows for Microservices-Style Applications [I]
46mThe End of Event-Based Architectures? Orchestration with Temporal