Global multi-region platform: request walkthrough and lab
Part of the Staff AI System Design course. Everything runs in your browser and makes no network requests. Every figure is illustrative.
Request walkthrough
- Responsible component
- Input
- Output and state
- Possible failure
Lab: When Is Recovery Eligible?
Is this recovery plan eligible?
Toy policy: A/B allowed in EU only, RPO 60 seconds, RTO 600 seconds, 20-second lease. Loss bound = gap at isolation + lease. The plan is detection 60 + fencing 30 + restore + route 30 + validation 60 seconds. Safety gates decide promotion; a plan over the RTO is a separate planning miss to rehearse or redesign, not a block. Policy, approval, audit and bundle gates are assumed true here; Python tests exercise them. Results are computed from assumptions, not measured.