Domain 4 of 4 · Chapter 3 of 4

Re-architecting migrated workloads

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing AWS Certified Solutions Architect - Professional premium course — no separate purchase.

Included in this chapter:

  • Choosing the target compute platform
  • Containerizing: ECS vs EKS, Fargate, App2Container
  • Database modernization: RDS, Aurora, SCT, Babelfish
  • Decomposing monoliths and storage re-architecture
  • Exam patterns: reading a re-architecture question right

Target compute model for a re-architected workload (SAP-C02)

DimensionEC2 (rehost/replatform)ECS on FargateEKSLambda
Operational ownershipOS + runtime + scaling you ownRuntime only; no hostsK8s control plane managed; you run workloadsNone (fully serverless)
Best fitCustom kernel, GPU, licensed AMI, lift-and-shiftStateless/containerized services, AWS-nativeExisting Kubernetes / portability needsShort event-driven functions
Hard limits to watchYou patch and scale the fleetTask CPU/memory sizingPer-cluster cost + K8s skills15-min run, 6 MB sync payload, 10 GB mem
Scaling modelAuto Scaling groupsService Auto Scaling per taskCluster Autoscaler / Karpenter / HPAAutomatic, per-request concurrency
Migration fitFastest target after rehostReplatform (containerize, drop hosts)Replatform if already on KubernetesRefactor (re-write to functions)

Decision tree

Fits a managed runtime? (no custom kernel/GPU/licensed AMI) No Amazon EC2 rehost/replatform; you patch Yes Short, event-driven, under 15 min? Yes AWS Lambda serverless functions No (containerize) Existing Kubernetes or portability need? No Amazon ECS on AWS Fargate lowest-ops containers Yes Amazon EKS K8s-conformant; Fargate/EC2 nodes Decouple with SQS / SNS / EventBridge / Step Functions; data tier to RDS, Aurora, or DynamoDB apply strangler-fig to break the monolith incrementally

Cheat sheet

  • Walk the managed-service ladder, stop at first fit
  • Lambda's hard limits disqualify long or heavy work
  • Fargate removes the host fleet; the EC2 launch type doesn't
  • ECS for lowest ops, EKS for existing Kubernetes
  • EKS is certified Kubernetes-conformant, so manifests run unchanged
  • Same image runs on ECS or EKS, pulled from ECR
  • App2Container containerizes .NET/Java without source code
  • Replatform a self-managed DB to RDS, refactor to Aurora
  • SCT converts the schema; DMS moves the data
  • Babelfish keeps SQL Server clients on Aurora PostgreSQL
  • Refactor to DynamoDB only for key-based access at scale
  • ElastiCache and OpenSearch are purpose-built offload targets
  • Strangler fig decomposes a monolith incrementally
  • An anti-corruption layer protects the new service model
  • Decouple services with SQS, SNS, EventBridge, or Step Functions
  • Step Functions: Standard exactly-once, Express at-least-once
  • Externalize state to S3, EFS, or FSx to make compute stateless
  • Use Aurora Serverless v2 (not v1) to scale down on idle
  • Re-architecting is the target design, not the move itself
  • Fargate task definitions require awsvpc, task-level CPU/memory, and an execution role
  • API Gateway REST usage plans give per-client throttling via API keys

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. AWS Lambda quotas
  2. Amazon ECS launch types (Fargate vs EC2)
  3. What is AWS Elastic Beanstalk
  4. What is Amazon EKS
  5. What is AWS App2Container
  6. What is Amazon RDS
  7. Amazon Aurora overview
  8. Aurora Global Database
  9. AWS Schema Conversion Tool (SCT)
  10. What is AWS Database Migration Service
  11. Babelfish for Aurora PostgreSQL
  12. Amazon DynamoDB introduction
  13. What is Amazon ElastiCache
  14. What is Amazon OpenSearch Service
  15. What is Amazon API Gateway
  16. What is Amazon SQS
  17. What is Amazon SNS
  18. What is Amazon EventBridge
  19. What is AWS Step Functions
  20. What is Amazon S3
  21. What is Amazon EFS
  22. Amazon FSx documentation
  23. Aurora Serverless v2