Resilience, Migration, and Business Continuity · Topic 2 of 6
Replatforming, Refactoring, and Legacy Modernization
Choose replatform or refactor on AWS: managed RDS, Aurora, ECS and Lambda, the strangler fig pattern, and AWS Transform for mainframe and .NET work.
- 5 min
- Replatform means managed services with minimal code change; refactor means changing the architecture.
- RDS and Aurora remove patching, backups and failover work for databases.
- ECS on Fargate is the lowest overhead container platform when no Kubernetes skills exist.
What it is
Replatforming changes the platform under an application without rewriting it. A self-managed database becomes Amazon RDS, or a server becomes a container on Amazon ECS, and AWS takes over patching, backups and failover.
Refactoring changes the application itself, for example splitting a monolith into microservices or moving work to AWS Lambda. Legacy modernization applies both ideas to old platforms such as mainframes and .NET Framework applications.
How it works
Replatforming with managed services
The goal is to remove undifferentiated operations work while changing as little code as possible.
| From | Replatform to | What you stop doing |
|---|---|---|
| Self-managed MySQL, PostgreSQL, SQL Server, Oracle | Amazon RDS | OS and engine patching, backups, standby management |
| Self-managed MySQL or PostgreSQL that needs scale | Amazon Aurora | Storage sizing, replica management, slow failover |
| Application on a VM | Amazon ECS on AWS Fargate | Host patching and capacity management |
| Scheduled scripts or small event handlers | AWS Lambda | Running servers for idle time |
| Self-managed message broker | Amazon MQ | Broker patching and high availability setup |
| Web app on a VM with standard runtime | AWS Elastic Beanstalk | Provisioning and load balancer setup |
Aurora keeps six copies of data across three Availability Zones and supports up to 15 Aurora Replicas. RDS Multi-AZ adds a standby with automatic failover. Both are typical answers when a scenario stresses "reduce operational overhead" for a database.
Containers are a common replatform target. If the team already runs Kubernetes, Amazon EKS keeps the tooling; if not, ECS with Fargate has the least to manage.
Refactoring and the strangler fig pattern
A big-bang rewrite of a monolith is risky. The strangler fig pattern replaces it piece by piece: a routing layer sends some requests to new services and the rest to the old system, until nothing is left in the monolith.
Amazon API Gateway, Application Load Balancer path rules and Amazon VPC Lattice can act as the routing layer. Each extracted service owns its own data store, and an anti-corruption layer can translate between the old data model and the new one.
Serverless adoption often uses events instead of direct calls. Amazon EventBridge, Amazon SQS and Amazon SNS decouple services, and AWS Step Functions coordinates multi-step workflows that used to live inside one process.
Mainframe and application modernization
AWS Transform uses AI agents to modernize IBM z/OS and Fujitsu GS21 mainframe applications. It analyzes the codebase, generates documentation, extracts business logic, decomposes the monolith into domains, plans waves and refactors COBOL workloads into Java applications.
People stay in the loop. AWS Transform asks for approval of intermediate results such as domain decomposition and wave planning, and it provides infrastructure as code templates for the target environment.
For Windows estates, AWS Transform modernizes .NET Framework applications to cross-platform .NET that runs on Linux. That removes Windows license cost and opens the door to containers on Linux.
| Legacy item | Modernization path |
|---|---|
| COBOL mainframe application | AWS Transform analysis, decomposition and refactor to Java |
| .NET Framework application | AWS Transform port to cross-platform .NET on Linux |
| Commercial database such as Oracle or SQL Server | Schema conversion plus AWS DMS to Aurora PostgreSQL or Aurora MySQL |
| Monolith on servers | Strangler fig extraction into containers or Lambda |
Choosing the right level
| Requirement in the question | Best fit |
|---|---|
| Fastest exit, no code change | Rehost |
| Less operations work, little code change | Replatform |
| Independent scaling and deployment of features | Refactor |
| Old language skills are scarce, logic must be kept | Mainframe modernization with AWS Transform |
How the exam asks about this
- "Move a self-managed PostgreSQL cluster with the LEAST operational overhead and minimal code change." -> Replatform to Amazon Aurora PostgreSQL or Amazon RDS for PostgreSQL.
- "A monolith must be modernized gradually with no big-bang cutover." -> Strangler fig pattern with API Gateway or ALB routing new paths to new services.
- "A COBOL application must be analyzed, documented and converted to Java while keeping business rules." -> AWS Transform for mainframe.
What to remember for the exam
- Replatform means managed services with minimal code change; refactor means changing the architecture.
- RDS and Aurora remove patching, backups and failover work for databases.
- ECS on Fargate is the lowest overhead container platform when no Kubernetes skills exist.
- The strangler fig pattern replaces a monolith incrementally behind a routing layer.
- AWS Transform covers mainframe refactoring and .NET Framework to cross-platform .NET.
- Engine changes for commercial databases pair schema conversion with AWS DMS.
"Refactor to microservices" is not the answer when the question asks for the fastest migration or minimal code change. Read the constraint first, then pick the lowest level of change that meets it.
Common mistakes
Picking Kubernetes for everything. EKS is right when the team already uses Kubernetes tooling, but ECS with Fargate has less to manage for a team new to containers.
Splitting services but sharing one database. Microservices that still share tables cannot be deployed or scaled independently, so each extracted service should own its data.
Rewriting a mainframe by hand with no analysis. Business logic is often undocumented. Analysis, documentation and decomposition come before code conversion.
Check yourself
What does the routing layer do in the strangler fig pattern?
It sends requests for extracted features to new services and all other requests to the monolith, so the old system shrinks gradually.
Which replatform target has the least container management for a team without Kubernetes experience?
Amazon ECS on AWS Fargate, because there are no hosts or control plane tooling to manage.
How can an organization cut Windows license cost for a .NET Framework application?
Use AWS Transform to port it to cross-platform .NET so it can run on Linux.
Topic 2 of 6 in Resilience, Migration, and Business Continuity
Test yourself on Resilience, Migration, and Business Continuity
The full AWS SAP-C03 question bank has 850 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by AWS. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.