The proven patterns that architects apply to build systems that are reliable, scalable, and maintainable.
Architecture patterns for the cloud:
1. Multi-tier Architecture
- Presentation tier (CDN/S3 + CloudFront)
- Application tier (EC2/Lambda behind ALB)
- Data tier (RDS Multi-AZ + ElastiCache)
- Each tier in separate subnets, security groups restrict lateral movement
2. Microservices
- Each service owns its data store
- Communicate via API Gateway, event bus, or service mesh
- Deploy independently, scale independently
- Service discovery: AWS Cloud Map, Kubernetes DNS
3. Event-Driven Architecture
- Producers emit events to EventBridge, SNS, or Kafka
- Consumers process asynchronously via Lambda, SQS, or ECS tasks
- Fan-out: SNS → multiple SQS queues
- Benefits: loose coupling, scalability, easy auditing
4. CQRS (Command Query Responsibility Segregation)
- Separate read and write models
- Write model: DynamoDB for writes (consistent)
- Read model: ElasticSearch or read replica for queries
5. Strangler Fig Pattern
- Incrementally migrate monolith to microservices
- New features built as services, old paths remain until replaced
6. Saga Pattern
- Distributed transactions across microservices
- Choreography (events) vs Orchestration (state machine/Step Functions)
7. Circuit Breaker
- Prevent cascading failures by detecting downstream errors
- States: Closed → Open → Half-Open
- AWS: use ALB health checks + Auto Scaling to replace unhealthy instances
1. Multi-tier Architecture
- Presentation tier (CDN/S3 + CloudFront)
- Application tier (EC2/Lambda behind ALB)
- Data tier (RDS Multi-AZ + ElastiCache)
- Each tier in separate subnets, security groups restrict lateral movement
2. Microservices
- Each service owns its data store
- Communicate via API Gateway, event bus, or service mesh
- Deploy independently, scale independently
- Service discovery: AWS Cloud Map, Kubernetes DNS
3. Event-Driven Architecture
- Producers emit events to EventBridge, SNS, or Kafka
- Consumers process asynchronously via Lambda, SQS, or ECS tasks
- Fan-out: SNS → multiple SQS queues
- Benefits: loose coupling, scalability, easy auditing
4. CQRS (Command Query Responsibility Segregation)
- Separate read and write models
- Write model: DynamoDB for writes (consistent)
- Read model: ElasticSearch or read replica for queries
5. Strangler Fig Pattern
- Incrementally migrate monolith to microservices
- New features built as services, old paths remain until replaced
6. Saga Pattern
- Distributed transactions across microservices
- Choreography (events) vs Orchestration (state machine/Step Functions)
7. Circuit Breaker
- Prevent cascading failures by detecting downstream errors
- States: Closed → Open → Half-Open
- AWS: use ALB health checks + Auto Scaling to replace unhealthy instances