Resiliency and Availability Design Patterns for the Cloud
summary
Cloud architectures use geographic redundancy (22 regions, multi-AZ with typically three AZs) plus queues, autoscaling and read/write separation to survive failures; DynamoDB hit 44 million TPS on Prime Day and S3 sustains millions/sec. Defensive code uses timeouts, exponential backoff with jitter and circuit breakers to avoid self‑DDOS, shuffle sharding (cells, e.g. 100 nodes with shard size 5) to cut blast radius toward zero, and chaos exercises (Netflix Simian Army / game days) to validate recoveries.
AI-written summary · about
chapters (13)
- 0:00Intro
- 3:26Partial failure mode
- 8:00Multi-AZ architecture
- 18:07Database Federation
- 18:28Database Sharding
- 27:11Circuit Breaker
- 28:52Probing for health
- 30:07Shallow health check
- 30:31Deep health check
- 38:28Cell-based architecture
- 39:52Shuffle sharding
- 44:15Chaos engineering
- 45:11Failure injection
related talks
57mResilient Cloud Architecture Design Patterns
20mMaximise database availability
31mDevOps in the Cloud: Case Studies of Amazon.com teams and their resilient architectures
25mChaos Engineering for Cloud native Apps
39mCloud Native Architectures
20mAmazon DynamoDB: A Scalable, Predictably Performant, and Fully Managed NoSQL...
35mBuilding Resilient Workloads on AWS: Tools and Techniques
26mThe tools and techniques behind resilient systems