Explore fault isolation boundaries on AWS
summary
AWS separates failure domains by Region, Availability Zone and PoP (31 Regions, 410+ PoPs) and splits services into control plane, data plane and zonal/regional/global types. Data plane operations (e.g., EC2 networking, DynamoDB reads/writes, Route 53 health checks) remain available when control plane actions (RunInstances, CreateRole, S3 bucket mapping in us-east-1) fail, so design for static stability: pre-provision resources, avoid control-plane steps in recovery paths, use CloudWatch for dimensioned observability and prefer data-plane recovery actions.
AI-written summary · about
chapters (7)
- 0:00Introduction to resilience
- 3:33Defining fault isolation
- 4:32AWS infrastructure boundaries
- 8:11Control vs. data planes
- 13:41Service categorization
- 18:40Architecting for recovery
- 22:05Recommendations and wrap-up
related talks
41mResiliency and data residency
49mPatterns for resilient, highly available cloud architectures
14mExpect the unexpected: Building resilience with AWS
48mResiliency and Availability Design Patterns for the Cloud
43mFlying Over AWS
35mDesign patterns for application resiliency
35mBuilding Resilient Workloads on AWS: Tools and Techniques
48mAprenda sobre la infraestructura Global de AWS