Beyond eleven nines - How Amazon S3 is built for durability
summary
Amazon S3 aims for eleven nines of durability by baking risk thinking into engineering culture via a formal durability review: a written threat model discussed in a ~two‑hour meeting with a 30‑minute silent read, then a tree of threats → risks → mitigations. Example: deep‑learning compression requires inline decompression validation, writing compressed objects to new files and keeping originals for weeks as a guardrail, and storing models in a versioned datastore. Observability uses metrics, canaries, auditors that continuously detect failures.
AI-written summary · about
chapters (4)
related talks
35mBuilding Systems that Last
19mCultural Foundation of Operational Excellence: Amazon’s Mechanisms for Resiliency
26mA proactive approach to resilience in modern apps
45mA resilient application in a partially degraded world
36mDesigning for failure - architecting fault-tolerant system
45m⚡A Resilient Application in a Partially Degraded World ⚡
01h 03mPatterns for Resilient Architecture
53mResiliency and Availability Design Patterns for the Cloud