system design

System Design: Multi-Region Architecture

Design a globally available application with multi-region deployment.

System Design: Multi-Region Architecture

The Problem

Design a global e-commerce platform that serves users in Europe, North America, and Asia with < 200ms latency. The system handles 10K requests/second and must survive an entire region going offline.

Architecture

Users → CloudFront/CDN → Regional ALBs → App Servers (ECS/K8s)
                                              ↓
                            Regional RDS (read replicas) + DynamoDB Global Tables
                                              ↓
                            S3 Cross-Region Replication (assets)

Key Decisions

Component	Choice	Why
DNS	Route 53 latency-based routing	Routes users to nearest healthy region
Database	DynamoDB Global Tables or Aurora Global Database	Multi-region active-active replication
Session	Stateless JWT + Redis Global	No sticky sessions needed
Assets	S3 + CloudFront	Edge-cached globally
Failover	Health checks + automatic DNS failover	Region failure → traffic reroutes in < 60s

Trade-offs

Consistency vs latency: eventual consistency across regions (acceptable for e-commerce)

Cost: 3x infrastructure cost vs single region

Complexity: data conflicts, clock skew, deployment coordination

Your design notes

Work through this problem yourself before reading the walkthrough above. Your notes are stored locally and not submitted anywhere — only sent to the AI when you click Review.