Beyond the State Machine: Scaling Order Orchestration
summary
Booking.com moves its Order Orchestration from AWS Step Functions to Temporal and proves specifics: a 75% cost reduction on a tested workflow, performance hitting 1,200 actions/sec with ~7.5s end-to-end happy‑path latency, and a code shrink from 2,000 YAML → 1,000 Java with a feature dev time dropping from two weeks to one day. The rollout uses an engine router with shadow mode for live comparison. Drawbacks: no built-in streaming, retry-by-exception gaps, extra work for AWS integrations, and initial memory leaks in activity code.
AI-written summary · about
chapters (21)
- 0:00Introduction and welcome
- 1:06Scale of Booking.com operations
- 2:09Era of static workflows
- 3:14Multi-product challenges
- 4:17Complexity of multi-product bundles
- 5:41Manageability wall with Step Functions
- 7:47Exploring Temporal for orchestration
- 9:14Conditional logic comparison
- 11:00Try-catch logic comparison
- 12:24Data manipulation costs
- 14:07Code-first approach benefits
- 15:11Migration feasibility questions
- 16:35Target architecture with Temporal
- 17:36Transitional architecture
- 19:02Evaluation results and drawbacks
- 21:28Order creation workflow reimplementation
- 23:35Cost and performance evaluation
- 24:58Performance testing challenges
- 26:23Temporal wins and pitfalls
- 28:08Choosing the right orchestration engine
- 28:50Summary and Q&A
related talks
44mGetting started with Orchestration and Choreography in Distributed Applications
50mThe end of event-based architectures? Orchestration with Temporal
47mComplex event flows in distributed systems
42mHow to build distributed systems in JavaScript
26mMicroservices Madness: Application And Infrastructure Automation Using Temporal
25mFinding Harmony in Distributed Systems: Choreography vs Orchestration in Software Dev
39mBuilding Scalable Orchestration
49mThe Benefits of Orchestration in Distributed Systems