Multi-Zone Service Architecture for 99.99% Uptime
summary
Claims 99.99% uptime by deploying services across multiple availability zones using Kubernetes and zone-aware PVs and pod affinity. Implements a metadata shard-assignment service that acts as an intelligent router and router/load balancer, uses consistent hashing to move partitions, and applies zone-aware autoscaling plus reserved instances for cost control. Chooses eventual consistency for writes, isolates stateful components, and runs CI/CD canaries. Outcome: service stays functional when a zone fails but incurs extra cost and operational complexity.
AI-written summary · about
chapters (10)
- 0:00Introduction and Speaker Background
- 1:22Importance of High Availability
- 1:51Microservices and Multi-Zone Architecture
- 2:44Foundational Principles of Multi-Zone Deployment
- 4:42Service Design for Multi-Zone Resilience
- 6:04Request Routing and Load Balancing
- 6:38Data Consistency and State Management
- 8:19Monitoring, Observability, and Cost Optimization
- 9:57Operational Best Practices
- 10:28Conclusion and Final Thoughts
related talks
25mStatic stability using availability zones
13mEnhancing Cloud Scalability and Fault Tolerance
33mMany Layers of Availability
45mKubernetes Resiliency
46mMeeting Impossible SLAs: How we made our uptime 99.999%
36mOptimizing Resource Usage in Kubernetes
31mHow We Went From Two Major Outages to 99.98% Reliability in 6 Months
19mZero-Downtime by Design: High-Availability Migration Patterns