| name | aws-patterns |
| description | AWS Patterns: Lambda, ECS/EKS, DynamoDB, S3, RDS, CloudWatch, serverless architecture. |
| triggers | {"extensions":[".py",".ts",".yaml"],"directories":["aws/","infrastructure/"],"keywords":["aws","lambda","dynamodb","s3","ecs","eks","cloudformation","terraform","sam","cdk","serverless"]} |
| auto_load_when | Building on AWS or designing cloud-native architectures |
| agent | cloud-architect |
| tools | ["Read","Write","Bash"] |
AWS Architecture Patterns
Focus: Serverless, containers, managed services, cost optimization
1. Compute Selection
When to use what:
├── Serverless (Lambda)
│ ├── Event-driven, intermittent workloads
│ ├── < 15 min execution, < 10GB memory
│ └── Low traffic (pay per invocation)
│
├── Containers (ECS/Fargate)
│ ├── Long-running services
│ ├── Need full control over runtime
│ └── Batch jobs, APIs, workers
│
├── Kubernetes (EKS)
│ ├── Multi-service architecture
│ ├── Need Kubernetes ecosystem
│ └── Complex orchestration needs
│
└── EC2
│ ├── Legacy workloads
│ ├── Specific hardware needs (GPU)
│ └── Full control required
2. Data Layer Patterns
Database selection:
├── DynamoDB (NoSQL)
│ ├── Massive scale, low latency
│ ├── Pay per request (good for variable load)
│ └── Single-table design for efficiency
│
├── Aurora (SQL)
│ ├── PostgreSQL/MySQL compatible
│ ├── Serverless option (pay per ACU)
│ └── Global tables for multi-region
│
├── RDS (SQL)
│ ├── Standard PostgreSQL/MySQL
│ ├── Less demanding than Aurora
│ └── Good for migrate-from-legacy
│
├── ElastiCache
│ ├── Redis for caching, sessions
│ └── Memcached for simple caching
│
└── S3 (Object storage)
├── Any file/blob storage
├── Lambda trigger on object create
└── Data lake with Athena
3. Event-Driven Architecture
AWS Event Patterns:
├── Lambda + S3
│ ├── Image processing, ETL triggers
│ └── Object created → Lambda invoked
│
├── Lambda + DynamoDB Streams
│ ├── Real-time changes captured
│ └── Audit logs, sync to other systems
│
├── EventBridge
│ ├── Decoupled microservices
│ ├── Schedule-based (cron)
│ └── Cross-account events
│
├── SQS + Lambda
│ ├── Durable queue for async work
│ └── Lambda polls queue (or SQS trigger)
│
└── SNS + Lambda
├── Pub/sub for notifications
└── Fan-out to multiple targets
4. Infrastructure as Code
IaC Tool Selection:
├── CloudFormation
│ ├── Native AWS, full coverage
│ └── JSON/YAML, template inheritance
│
├── CDK (Cloud Development Kit)
│ ├── Code in TypeScript/Python
│ └── Generates CloudFormation
│ └── Best of both worlds
│
├── Serverless Application Model (SAM)
│ ├── Simplified for serverless
│ └── Local testing with sam local
│
└── Terraform
├── Multi-cloud (AWS + others)
└── State management needed
5. Cost Optimization
AWS Cost Patterns:
├── Compute
│ ├── Lambda: Pay per invocation
│ ├── Fargate: Pay per vCPU/GB-hour
│ ├── Reserved Instances: 1-3 year for stable
│ └── Spot: Batch jobs, stateless workers (70%+ off)
│
├── Storage
│ ├── S3 Intelligent-Tiering (auto-move)
│ ├── Glacier for archives
│ └── Lifecycle policies
│
├── Database
│ ├── Aurora Serverless for variable
│ └── Provisioned for steady state
│
└── Networking
├── CloudFront for static assets
└── PrivateLink for VPC-to-VPC
Key Patterns
- Serverless-first - Start with Lambda, add complexity only if needed
- Single table DynamoDB - Design for access patterns upfront
- Event-driven - Decouple services with EventBridge/SNS/SQS
- Infrastructure as code - Never click in console for prod
- Cost-aware - Use Spot, Savings Plans, S3 Intelligent-Tiering
Anti-Patterns
❌ Using EC2 for everything — overprovisioned, underutilized
✅ Start with Lambda, move to containers only if needed
❌ No caching — hitting database for every request
✅ ElastiCache (Redis) for session/cache, CloudFront for static
❌ No dead letter queue — failed messages lost
✅ Configure DLQ on SQS/Lambda failures
❌ Single-AZ deployment — one AZ outage = downtime
✅ Multi-AZ for all production workloads
❌ No cost monitoring — surprised by bill at end of month
✅ Budget alerts, cost explorer, rightsizing recommendations
Quick Reference
| Service | Use Case | Key Feature |
|---|
| Lambda | Serverless functions | Pay per invocation |
| DynamoDB | NoSQL, massive scale | Single-digit ms |
| S3 | Object storage | Intelligent-Tiering |
| RDS/Aurora | Relational | Managed backups |
| ElastiCache | Caching | Redis/Memcached |
| CloudFront | CDN | Global edge |
| EventBridge | Event bus | Decoupling |
| API Gateway | REST APIs | Rate limiting |
🌍 Universal Language Support
- Turkish Native: This skill natively supports Turkish. If the user prompt is in Turkish, all analysis, formatting, and output MUST be entirely in Turkish. You do not need explicit "write in Turkish" instructions.
